diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml
new file mode 100644
index 0000000..a07850d
--- /dev/null
+++ b/.github/ISSUE_TEMPLATE/bug_report.yml
@@ -0,0 +1,81 @@
+name: Bug report
+description: Report a reproducible defect in the MCP server or its contracts.
+title: "bug: "
+body:
+  - type: markdown
+    attributes:
+      value: >-
+        Do not include secrets or undisclosed vulnerabilities. Use SECURITY.md
+        for private security reports.
+  - type: input
+    id: version
+    attributes:
+      label: Version or commit
+      description: Include the exact crate version or Git commit.
+      placeholder: 0.2.0 or abc1234
+    validations:
+      required: true
+  - type: dropdown
+    id: tool_pack
+    attributes:
+      label: Tool pack
+      options:
+        - core (default, 13 tools)
+        - standard (33 tools)
+        - full (47 tools)
+        - unknown
+    validations:
+      required: true
+  - type: input
+    id: client_protocol
+    attributes:
+      label: Client and negotiated protocol
+      placeholder: Codex CLI; 2025-11-25
+    validations:
+      required: true
+  - type: textarea
+    id: reproduction
+    attributes:
+      label: Minimal reproduction
+      description: Provide ordered steps and a minimal, redacted request.
+    validations:
+      required: true
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected result
+    validations:
+      required: true
+  - type: textarea
+    id: actual
+    attributes:
+      label: Actual result
+    validations:
+      required: true
+  - type: dropdown
+    id: contract_impact
+    attributes:
+      label: Contract impact
+      options:
+        - Incorrect success or fail-open behavior
+        - Tool discovery or schema mismatch
+        - Persistence or audit mismatch
+        - Startup or transport failure
+        - Other
+    validations:
+      required: true
+  - type: textarea
+    id: logs
+    attributes:
+      label: Redacted diagnostics
+      description: Include stderr only; remove secrets and personal paths.
+      render: text
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Preflight
+      options:
+        - label: I reproduced this on the stated version or commit.
+          required: true
+        - label: I removed secrets and unrelated private data.
+          required: true
diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml
new file mode 100644
index 0000000..3ba13e0
--- /dev/null
+++ b/.github/ISSUE_TEMPLATE/config.yml
@@ -0,0 +1 @@
+blank_issues_enabled: false
diff --git a/.github/ISSUE_TEMPLATE/feature_request.yml b/.github/ISSUE_TEMPLATE/feature_request.yml
new file mode 100644
index 0000000..ec1eedc
--- /dev/null
+++ b/.github/ISSUE_TEMPLATE/feature_request.yml
@@ -0,0 +1,55 @@
+name: Feature request
+description: Propose a user outcome or auditable contract improvement.
+title: "feat: "
+body:
+  - type: textarea
+    id: problem
+    attributes:
+      label: Problem
+      description: What user problem cannot be solved with the current contract?
+    validations:
+      required: true
+  - type: textarea
+    id: outcome
+    attributes:
+      label: Desired user flow
+      description: Show the smallest useful request-to-audit path.
+    validations:
+      required: true
+  - type: textarea
+    id: evidence
+    attributes:
+      label: Evidence and failure boundary
+      description: What must be supplied, and when must the server stop or fail closed?
+    validations:
+      required: true
+  - type: dropdown
+    id: tool_pack
+    attributes:
+      label: Narrowest justified tool pack
+      options:
+        - core
+        - standard
+        - full compatibility only
+        - no new tool
+        - unsure
+    validations:
+      required: true
+  - type: textarea
+    id: alternatives
+    attributes:
+      label: Existing tools and alternatives
+      description: Explain why composition of current tools is insufficient.
+    validations:
+      required: true
+  - type: checkboxes
+    id: checks
+    attributes:
+      label: Contract checks
+      options:
+        - label: The proposal does not claim access to hidden chain of thought.
+          required: true
+        - label: Model-backed or retrieval-backed behavior has an explicit provider boundary.
+          required: true
+        - label: I checked features/manifest.json for an overlapping planning record.
+          required: true
diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
new file mode 100644
index 0000000..26dcf02
--- /dev/null
+++ b/.github/pull_request_template.md
@@ -0,0 +1,21 @@
+## Outcome
+
+Describe the user-visible result and why the change is needed.
+
+## Contract impact
+
+- [ ] No public MCP behavior changes.
+- [ ] Tool-pack inventory and `evals/tool-packs.json` are synchronized.
+- [ ] Evidence, provider, and fail-closed boundaries remain explicit.
+- [ ] Persistence, audit, and compatibility effects are documented.
+
+## Verification
+
+List exact commands and results. For behavior changes, include the failing test
+observed before the implementation and its passing result afterward.
+
+- [ ] `just check`
+- [ ] `just eval` when protocol, schema, routing, or governance changed
+- [ ] `just audit` when dependencies changed
+- [ ] Documentation and examples match the implemented behavior
+- [ ] No secrets, local absolute paths, generated caches, or private artifacts
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 61d922b..1dd3e18 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -6,6 +6,9 @@ on:
   pull_request:
     branches: [main]
 
+permissions:
+  contents: read
+
 jobs:
   reasonkit-think-mcp:
     runs-on: ubuntu-latest
@@ -15,11 +18,34 @@ jobs:
         run: |
           rustup toolchain install 1.95.0 --profile minimal --component rustfmt,clippy
           rustup default 1.95.0
+      - name: Install uv 0.12.1
+        uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
+        with:
+          version: "0.12.1"
+          enable-cache: false
       - name: Format
         run: cargo fmt --check
       - name: Clippy
-        run: cargo clippy -- -D warnings
+        run: cargo clippy --all-targets --all-features -- -D warnings
+      - name: Rust unit and integration tests
+        run: cargo test --all-targets --all-features
+      - name: Contract unit tests
+        run: python3 -m unittest discover -s tests -p "test_*.py" -v
+      - name: Package release candidate
+        run: CARGO_TARGET_DIR=target/package-check cargo package --locked
+      - name: Validate MCP Registry schema
+        run: uvx --from check-jsonschema==0.38.0 check-jsonschema --schemafile https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json server.json
+      - name: Install cargo-dist 0.32.0
+        run: cargo install cargo-dist --version 0.32.0 --locked
+      - name: Validate configuration-only cargo-dist manifest
+        run: dist manifest --artifacts=local --output-format=json --no-local-paths
       - name: Build release
         run: cargo build --release
-      - name: Smoke test (stdio MCP)
+      - name: Full-pack smoke test (stdio MCP)
         run: python3 scripts/smoke_test.py
+      - name: MCP contract evaluations
+        run: python3 evals/run_contract_evals.py --binary target/release/reasonkit-think-mcp
+      - name: Install cargo-audit
+        run: cargo install cargo-audit --locked
+      - name: Dependency security audit
+        run: cargo audit --deny warnings
diff --git a/.gitignore b/.gitignore
index d1b1e97..1663837 100644
--- a/.gitignore
+++ b/.gitignore
@@ -4,4 +4,5 @@
 .env.*
 /target_check/
 /local-target/
-scripts/__pycache__/
+**/__pycache__/
+*.py[cod]
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4de842c..66d2bb8 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,8 +2,21 @@
 
 All notable changes to ReasonKit Think are documented here.
 
-## Unreleased
+## Unreleased — 0.2.0 release candidate
 
+- Added explicit `--version`, `--help`, and secret-free `doctor --json`
+  diagnostics, including fail-closed backend validation, without changing the
+  no-argument MCP stdio contract.
+- Upgraded the official Rust MCP SDK from `rmcp` 1.6.0 to 3.1.4 while keeping
+  the server stdio-only and extending negotiation checks through MCP 2026-07-28.
+- Added startup-fixed `core` (13, default), `standard` (33), and `full` (47)
+  tool packs with a full-pack compatibility rollback and pack-aware guidance.
+- Added `reasoning://config/tool-packs` and dual text plus
+  `structuredContent` JSON for successful results and handler-generated
+  application errors.
+- Removed the unbound `reasonkit-mem` cache dependency. The full-pack
+  `think_query` compatibility route now fails closed with
+  `experimental_memory_unavailable` instead of simulating empty retrieval.
 - Added `run_thinking_mode` as the recommended user-facing entry point for Auto,
   Quick, Explore, Map, Sketch, and Test workflows.
 - Added native MCP resource discovery/read support for `reasoning://thinking-modes`
@@ -19,3 +32,20 @@ All notable changes to ReasonKit Think are documented here.
   `get_thinking_history` with `deliberation_id`, and boolean-like audit flags.
 - Removed default pseudo-evidence from Auto/autopilot verification; unsupported
   claims remain `DATA_DEFICIT` until caller evidence is supplied.
+- Preserved verification blockers across repeated verification calls, prevented
+  assumptions from self-verifying without evidence, enforced ReAct action and
+  observation transitions, and rejected unknown pipeline stages.
+- Made verification criticality monotonic across normalized-claim upserts,
+  made assumption `critical: true` authoritative over conflicting lower
+  criticality, and kept pipeline and consensus policy overrides call-scoped.
+- Required an agent-authored frontier for consensus, excluded heuristic
+  scaffold/root nodes from allowed answers, and blocked failed pipeline gates
+  and unresolved checkpoints. Authorship is provenance- and node-type-backed;
+  reserved-tag spoofing is rejected, and answer provenance lists only selected
+  candidates. Blocked MCP responses use the exact `consensus_blocked` code and
+  omit the candidate final answer.
+- Added deterministic canonical JSON and PR-ready Markdown reasoning audits with
+  consistent artifact IDs and neutralized caller-controlled Markdown content.
+- Added deterministic protocol, pack, eight-check adversarial governance,
+  packaging, Registry, cargo-dist, and dependency-audit gates for CI and local
+  release checks.
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
new file mode 100644
index 0000000..026b266
--- /dev/null
+++ b/CONTRIBUTING.md
@@ -0,0 +1,44 @@
+# Contributing to ReasonKit Think
+
+ReasonKit Think is an auditable stdio MCP server. Contributions should preserve
+its fail-closed evidence boundaries, deterministic contracts, and small default
+tool surface.
+
+## Set up
+
+Install Rust 1.95, Python 3, `uv`, and `just`, then work from a source checkout:
+
+```sh
+cargo build --locked
+just check
+```
+
+`just check` runs formatting, Clippy, Rust tests, and Python contract tests. Run
+`just eval` for protocol, schema, tool, routing, or governance changes. Run
+`just audit` for dependency changes; the zero-warning baseline and prior cleanup
+are documented in [`docs/dependency-audit.md`](docs/dependency-audit.md).
+For release-metadata changes, run `just package-check` and `just registry-check`.
+If editing `dist-workspace.toml`, install cargo-dist 0.32.0 with
+`cargo install cargo-dist --version 0.32.0 --locked`, then run `just dist-check`.
+
+## Make a focused change
+
+1. Open an issue for behavior that changes a public tool, schema, or policy.
+2. Add a failing test before changing behavior.
+3. Keep stdout reserved for MCP frames; diagnostics belong on stderr.
+4. Update documentation and contract fixtures with the implementation.
+5. Use Conventional Commits, for example `fix(protocol): reject unknown stages`.
+
+Tool discovery is pack-aware. Any tool-surface change must update
+[`evals/tool-packs.json`](evals/tool-packs.json) and preserve exact lexical
+counts for every affected pack. New tools should land in the narrowest justified
+pack; `full` is the compatibility escape hatch, not the default.
+
+## Pull-request evidence
+
+Describe the user-visible outcome, contract or security impact, tests run, and
+any known limitation. Include the exact failing-then-passing test for behavior
+changes. Do not commit secrets, local absolute paths, generated caches, private
+planning material, or release automation that publishes on a tag.
+
+By contributing, you agree that your work is licensed under Apache-2.0.
diff --git a/Cargo.lock b/Cargo.lock
index 3d87150..0efe88f 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -2,12 +2,6 @@
 # It is not intended for manual editing.
 version = 4
 
-[[package]]
-name = "adler2"
-version = "2.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "320119579fcad9c21884f5c4861d16174d0e06250625266f50fe6898340abefa"
-
 [[package]]
 name = "aho-corasick"
 version = "1.1.4"
@@ -17,12 +11,6 @@ dependencies = [
  "memchr",
 ]
 
-[[package]]
-name = "allocator-api2"
-version = "0.2.21"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "683d7910e743518b0e34f1186f92494becacb047c7b6bf616c96772180fef923"
-
 [[package]]
 name = "android_system_properties"
 version = "0.1.5"
@@ -32,112 +20,11 @@ dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "anstream"
-version = "1.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "824a212faf96e9acacdbd09febd34438f8f711fb84e09a8916013cd7815ca28d"
-dependencies = [
- "anstyle",
- "anstyle-parse",
- "anstyle-query",
- "anstyle-wincon",
- "colorchoice",
- "is_terminal_polyfill",
- "utf8parse",
-]
-
-[[package]]
-name = "anstyle"
-version = "1.0.14"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "940b3a0ca603d1eade50a4846a2afffd5ef57a9feac2c0e2ec2e14f9ead76000"
-
-[[package]]
-name = "anstyle-parse"
-version = "1.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "52ce7f38b242319f7cabaa6813055467063ecdc9d355bbb4ce0c68908cd8130e"
-dependencies = [
- "utf8parse",
-]
-
-[[package]]
-name = "anstyle-query"
-version = "1.1.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "40c48f72fd53cd289104fc64099abca73db4166ad86ea0b4341abe65af83dadc"
-dependencies = [
- "windows-sys 0.61.2",
-]
-
-[[package]]
-name = "anstyle-wincon"
-version = "3.0.11"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "291e6a250ff86cd4a820112fb8898808a366d8f9f58ce16d1f538353ad55747d"
-dependencies = [
- "anstyle",
- "once_cell_polyfill",
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "anyhow"
-version = "1.0.102"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7f202df86484c868dbad7eaa557ef785d5c66295e41b460ef922eca0723b842c"
-
-[[package]]
-name = "arc-swap"
-version = "1.9.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6a3a1fd6f75306b68087b831f025c712524bcb19aad54e557b1129cfa0a2b207"
-dependencies = [
- "rustversion",
-]
-
-[[package]]
-name = "async-stream"
-version = "0.3.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0b5a71a6f37880a80d1d7f19efd781e4b5de42c88f0722cc13bcb6cc2cfe8476"
-dependencies = [
- "async-stream-impl",
- "futures-core",
- "pin-project-lite",
-]
-
-[[package]]
-name = "async-stream-impl"
-version = "0.3.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c7c24de15d275a1ecfd47a380fb4d5ec9bfe0933f309ed5e705b775596a3574d"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "async-trait"
-version = "0.1.89"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9035ad2d096bed7955a320ee7e2230574d28fd3c3a0f186cbea1ff3c7eed5dbb"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "atomic-polyfill"
-version = "1.0.3"
+version = "1.0.104"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8cf2bce30dfe09ef0bfaef228b9d414faaf7e563035494d7fe092dba54b300f4"
-dependencies = [
- "critical-section",
-]
+checksum = "330a5ed07fa54e4702c9d6c4174f74427fc0ef6e214bbd677ae50a5099946470"
 
 [[package]]
 name = "atomic-waker"
@@ -151,126 +38,24 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c08606f8c3cbf4ce6ec8e28fb0014a2c086708fe954eaa885384a6165172e7e8"
 
-[[package]]
-name = "axum"
-version = "0.7.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "edca88bc138befd0323b20752846e6587272d3b03b0343c8ea28a6f819e6e71f"
-dependencies = [
- "async-trait",
- "axum-core",
- "bytes",
- "futures-util",
- "http",
- "http-body",
- "http-body-util",
- "itoa",
- "matchit",
- "memchr",
- "mime",
- "percent-encoding",
- "pin-project-lite",
- "rustversion",
- "serde",
- "sync_wrapper",
- "tower 0.5.3",
- "tower-layer",
- "tower-service",
-]
-
-[[package]]
-name = "axum-core"
-version = "0.4.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "09f2bd6146b97ae3359fa0cc6d6b376d9539582c7b4220f041a33ec24c226199"
-dependencies = [
- "async-trait",
- "bytes",
- "futures-util",
- "http",
- "http-body",
- "http-body-util",
- "mime",
- "pin-project-lite",
- "rustversion",
- "sync_wrapper",
- "tower-layer",
- "tower-service",
-]
-
 [[package]]
 name = "base64"
 version = "0.22.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "72b3254f16251a8381aa12e40e3c4d2f0199f8c6508fbecb9d91f575e0fbb8c6"
 
-[[package]]
-name = "bitflags"
-version = "1.3.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
-
 [[package]]
 name = "bitflags"
 version = "2.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c4512299f36f043ab09a583e57bceb5a5aab7a73db1805848e8fef3c9e8c78b3"
 
-[[package]]
-name = "bitpacking"
-version = "0.9.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "96a7139abd3d9cebf8cd6f920a389cf3dc9576172e32f4563f188cae3c3eb019"
-dependencies = [
- "crunchy",
-]
-
-[[package]]
-name = "block-buffer"
-version = "0.10.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3078c7629b62d3f0439517fa394996acacc5cbc91c5a20d8c658e77abd503a71"
-dependencies = [
- "generic-array",
-]
-
-[[package]]
-name = "bon"
-version = "3.9.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f47dbe92550676ee653353c310dfb9cf6ba17ee70396e1f7cf0a2020ad49b2fe"
-dependencies = [
- "bon-macros",
- "rustversion",
-]
-
-[[package]]
-name = "bon-macros"
-version = "3.9.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "519bd3116aeeb42d5372c29d982d16d0170d3d4a5ed85fc7dd91642ffff3c67c"
-dependencies = [
- "darling 0.23.0",
- "ident_case",
- "prettyplease",
- "proc-macro2",
- "quote",
- "rustversion",
- "syn",
-]
-
 [[package]]
 name = "bumpalo"
 version = "3.20.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5d20789868f4b01b2f2caec9f5c4e0213b41e3e5702a50157d699ae31ced2fcb"
 
-[[package]]
-name = "byteorder"
-version = "1.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
-
 [[package]]
 name = "bytes"
 version = "1.11.1"
@@ -284,17 +69,9 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d16d90359e986641506914ba71350897565610e87ce0ad9e6f28569db3dd5c6d"
 dependencies = [
  "find-msvc-tools",
- "jobserver",
- "libc",
  "shlex",
 ]
 
-[[package]]
-name = "census"
-version = "0.4.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4f4c707c6a209cbe82d10abd08e1ea8995e9ea937d2550646e02798948992be0"
-
 [[package]]
 name = "cfg-if"
 version = "1.0.4"
@@ -307,6 +84,17 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
 
+[[package]]
+name = "chacha20"
+version = "0.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d524456ba66e72eb8b115ff89e01e497f8e6d11d78b70b1aa13c0fbd97540a81"
+dependencies = [
+ "cfg-if",
+ "cpufeatures",
+ "rand_core",
+]
+
 [[package]]
 name = "chrono"
 version = "0.4.44"
@@ -321,81 +109,6 @@ dependencies = [
  "windows-link",
 ]
 
-[[package]]
-name = "clap"
-version = "4.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1ddb117e43bbf7dacf0a4190fef4d345b9bad68dfc649cb349e7d17d28428e51"
-dependencies = [
- "clap_builder",
- "clap_derive",
-]
-
-[[package]]
-name = "clap_builder"
-version = "4.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "714a53001bf66416adb0e2ef5ac857140e7dc3a0c48fb28b2f10762fc4b5069f"
-dependencies = [
- "anstream",
- "anstyle",
- "clap_lex",
- "strsim",
-]
-
-[[package]]
-name = "clap_derive"
-version = "4.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f2ce8604710f6733aa641a2b3731eaa1e8b3d9973d5e3565da11800813f997a9"
-dependencies = [
- "heck",
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "clap_lex"
-version = "1.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8d4a3bb8b1e0c1050499d1815f5ab16d04f0959b233085fb31653fbfc9d98f9"
-
-[[package]]
-name = "cobs"
-version = "0.3.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0fa961b519f0b462e3a3b4a34b64d119eeaca1d59af726fe450bbba07a9fc0a1"
-dependencies = [
- "thiserror 2.0.18",
-]
-
-[[package]]
-name = "colorchoice"
-version = "1.0.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1d07550c9036bf2ae0c684c4297d503f838287c83c53686d05370d0e139ae570"
-
-[[package]]
-name = "core-foundation"
-version = "0.9.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91e195e091a93c46f7102ec7818a2aa394e1e1771c3ab4825963fa03e45afb8f"
-dependencies = [
- "core-foundation-sys",
- "libc",
-]
-
-[[package]]
-name = "core-foundation"
-version = "0.10.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b2a6cd9ae233e7f62ba4e9353e81a88df7fc8a5987b8d445b4d90c879bd156f6"
-dependencies = [
- "core-foundation-sys",
- "libc",
-]
-
 [[package]]
 name = "core-foundation-sys"
 version = "0.8.7"
@@ -404,209 +117,45 @@ checksum = "773648b94d0e5d620f64f280777445740e61fe701025087ec8b57f45c791888b"
 
 [[package]]
 name = "cpufeatures"
-version = "0.2.17"
+version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "59ed5838eebb26a2bb2e58f6d5b5316989ae9d08bab10e0e6d103e656d1b0280"
+checksum = "8b2a41393f66f16b0823bb79094d54ac5fbd34ab292ddafb9a0456ac9f87d201"
 dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "crc32fast"
-version = "1.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9481c1c90cbf2ac953f07c8d4a58aa3945c425b7185c9154d67a65e4230da511"
-dependencies = [
- "cfg-if",
-]
-
-[[package]]
-name = "critical-section"
-version = "1.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "790eea4361631c5e7d22598ecd5723ff611904e3344ce8720784c93e3d83d40b"
-
-[[package]]
-name = "crossbeam-channel"
-version = "0.5.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "82b8f8f868b36967f9606790d1903570de9ceaf870a7bf9fbbd3016d636a2cb2"
-dependencies = [
- "crossbeam-utils",
-]
-
-[[package]]
-name = "crossbeam-deque"
-version = "0.8.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9dd111b7b7f7d55b72c0a6ae361660ee5853c9af73f70c3c2ef6858b950e2e51"
-dependencies = [
- "crossbeam-epoch",
- "crossbeam-utils",
-]
-
-[[package]]
-name = "crossbeam-epoch"
-version = "0.9.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5b82ac4a3c2ca9c3460964f020e1402edd5753411d7737aa39c3714ad1b5420e"
-dependencies = [
- "crossbeam-utils",
-]
-
-[[package]]
-name = "crossbeam-utils"
-version = "0.8.21"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d0a5c400df2834b80a4c3327b3aad3a4c4cd4de0629063962b03235697506a28"
-
-[[package]]
-name = "crunchy"
-version = "0.2.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "460fbee9c2c2f33933d720630a6a0bac33ba7053db5344fac858d4b8952d77d5"
-
-[[package]]
-name = "crypto-common"
-version = "0.1.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "78c8292055d1c1df0cce5d180393dc8cce0abec0a7102adb6c7b1eef6016d60a"
-dependencies = [
- "generic-array",
- "typenum",
-]
-
 [[package]]
 name = "darling"
-version = "0.20.11"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fc7f46116c46ff9ab3eb1597a45688b6715c6e628b5c133e288e709a29bcb4ee"
-dependencies = [
- "darling_core 0.20.11",
- "darling_macro 0.20.11",
-]
-
-[[package]]
-name = "darling"
-version = "0.23.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "25ae13da2f202d56bd7f91c25fba009e7717a1e4a1cc98a76d844b65ae912e9d"
-dependencies = [
- "darling_core 0.23.0",
- "darling_macro 0.23.0",
-]
-
-[[package]]
-name = "darling_core"
-version = "0.20.11"
+version = "0.24.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0d00b9596d185e565c2207a0b01f8bd1a135483d02d9b7b0a54b11da8d53412e"
+checksum = "ed17f5901b6630b993ca003def43f2f8ef4014fc13b047b57aad617ff32bc2ec"
 dependencies = [
- "fnv",
- "ident_case",
- "proc-macro2",
- "quote",
- "strsim",
- "syn",
+ "darling_core",
+ "darling_macro",
 ]
 
 [[package]]
 name = "darling_core"
-version = "0.23.0"
+version = "0.24.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9865a50f7c335f53564bb694ef660825eb8610e0a53d3e11bf1b0d3df31e03b0"
+checksum = "6837e2cf7485aaae18f86181d2f0e9a7ed297a025e220aeabf63fdebd3a2ddff"
 dependencies = [
  "ident_case",
  "proc-macro2",
  "quote",
  "strsim",
- "syn",
+ "syn 3.0.3",
 ]
 
 [[package]]
 name = "darling_macro"
-version = "0.20.11"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fc34b93ccb385b40dc71c6fceac4b2ad23662c7eeb248cf10d529b7e055b6ead"
-dependencies = [
- "darling_core 0.20.11",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "darling_macro"
-version = "0.23.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ac3984ec7bd6cfa798e62b4a642426a5be0e68f9401cfc2a01e3fa9ea2fcdb8d"
-dependencies = [
- "darling_core 0.23.0",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "dashmap"
-version = "5.5.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "978747c1d849a7d2ee5e8adc0159961c48fb7e5db2f06af6723b80123bb53856"
-dependencies = [
- "cfg-if",
- "hashbrown 0.14.5",
- "lock_api",
- "once_cell",
- "parking_lot_core 0.9.12",
-]
-
-[[package]]
-name = "deranged"
-version = "0.5.8"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7cd812cc2bc1d69d4764bd80df88b4317eaef9e773c75226407d9bc0876b211c"
-dependencies = [
- "powerfmt",
- "serde_core",
-]
-
-[[package]]
-name = "derive_builder"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "507dfb09ea8b7fa618fcf76e953f4f5e192547945816d5358edffe39f6f94947"
-dependencies = [
- "derive_builder_macro",
-]
-
-[[package]]
-name = "derive_builder_core"
-version = "0.20.2"
+version = "0.24.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2d5bcf7b024d6835cfb3d473887cd966994907effbe9227e8c8219824d06c4e8"
+checksum = "2ac7135c3ef02b2f7833bbeb1be5ba7f966dcde8a87c6b87f65a778d71a02785"
 dependencies = [
- "darling 0.20.11",
- "proc-macro2",
+ "darling_core",
  "quote",
- "syn",
-]
-
-[[package]]
-name = "derive_builder_macro"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab63b0e2bf4d5928aff72e83a7dace85d7bba5fe12dcc3c5a572d78caffd3f3c"
-dependencies = [
- "derive_builder_core",
- "syn",
-]
-
-[[package]]
-name = "digest"
-version = "0.10.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
-dependencies = [
- "block-buffer",
- "crypto-common",
+ "syn 3.0.3",
 ]
 
 [[package]]
@@ -615,28 +164,7 @@ version = "6.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "16f5094c54661b38d03bd7e50df373292118db60b585c08a411c6d840017fe7d"
 dependencies = [
- "dirs-sys 0.5.0",
-]
-
-[[package]]
-name = "dirs"
-version = "5.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "44c45a9d03d6676652bcb5e724c7e988de1acad23a711b5217ab9cbecbec2225"
-dependencies = [
- "dirs-sys 0.4.1",
-]
-
-[[package]]
-name = "dirs-sys"
-version = "0.4.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "520f05a5cbd335fae5a99ff7a6ab8627577660ee5cfd6a94a6a929b52ff0321c"
-dependencies = [
- "libc",
- "option-ext",
- "redox_users 0.4.6",
- "windows-sys 0.48.0",
+ "dirs-sys",
 ]
 
 [[package]]
@@ -647,8 +175,8 @@ checksum = "e01a3366d27ee9890022452ee61b2b63a67e6f13f58900b651ff5665f0bb1fab"
 dependencies = [
  "libc",
  "option-ext",
- "redox_users 0.5.2",
- "windows-sys 0.61.2",
+ "redox_users",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -659,119 +187,33 @@ checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
-[[package]]
-name = "downcast-rs"
-version = "2.0.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "117240f60069e65410b3ae1bb213295bd828f707b5bec6596a1afc8793ce0cbc"
-
 [[package]]
 name = "dyn-clone"
 version = "1.0.20"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d0881ea181b1df73ff77ffaaf9c7544ecc11e82fba9b5f27b262a3c73a332555"
 
-[[package]]
-name = "either"
-version = "1.15.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "48c757948c5ede0e46177b7add2e67155f70e33c07fea8284df6576da70b3719"
-
-[[package]]
-name = "embedded-io"
-version = "0.4.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ef1a6892d9eef45c8fa6b9e0086428a2cca8491aca8f787c534a3d6d0bcb3ced"
-
-[[package]]
-name = "embedded-io"
-version = "0.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "edd0f118536f44f5ccd48bcb8b111bdc3de888b58c74639dfb034a357d0f206d"
-
-[[package]]
-name = "encoding_rs"
-version = "0.8.35"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "75030f3c4f45dafd7586dd6780965a8c7e8e285a5ecb86713e63a79c5b2766f3"
-dependencies = [
- "cfg-if",
-]
-
 [[package]]
 name = "equivalent"
 version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "877a4ace8713b0bcf2a4e7eec82529c029f1d0619886d18145fea96c3ffe5c0f"
 
-[[package]]
-name = "errno"
-version = "0.3.14"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
-dependencies = [
- "libc",
- "windows-sys 0.61.2",
-]
-
-[[package]]
-name = "fastdivide"
-version = "0.4.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9afc2bd4d5a73106dd53d10d73d3401c2f32730ba2c0b93ddb888a8983680471"
-
-[[package]]
-name = "fastrand"
-version = "2.4.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9f1f227452a390804cdb637b74a86990f2a7d7ba4b7d5693aac9b4dd6defd8d6"
-
 [[package]]
 name = "find-msvc-tools"
 version = "0.1.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5baebc0774151f905a1a2cc41989300b1e6fbb29aff0ceffa1064fdd3088d582"
 
-[[package]]
-name = "flate2"
-version = "1.1.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "843fba2746e448b37e26a819579957415c8cef339bf08564fe8b7ddbd959573c"
-dependencies = [
- "crc32fast",
- "miniz_oxide",
-]
-
-[[package]]
-name = "fnv"
-version = "1.0.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
-
 [[package]]
 name = "foldhash"
 version = "0.1.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"
 
-[[package]]
-name = "foreign-types"
-version = "0.3.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f6f339eb8adc052cd2ca78910fda869aefa38d22d5cb648e6485e4d3fc06f3b1"
-dependencies = [
- "foreign-types-shared",
-]
-
-[[package]]
-name = "foreign-types-shared"
-version = "0.1.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "00b0228411908ca8685dba7fc2cdd70ec9990a6e753e89b6ac91a84c40fbaf4b"
-
 [[package]]
 name = "form_urlencoded"
 version = "1.2.2"
@@ -781,26 +223,6 @@ dependencies = [
  "percent-encoding",
 ]
 
-[[package]]
-name = "fs2"
-version = "0.4.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9564fc758e15025b46aa6643b1b77d047d1a56a1aea6e01002ac0c7026876213"
-dependencies = [
- "libc",
- "winapi",
-]
-
-[[package]]
-name = "fs4"
-version = "0.13.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8640e34b88f7652208ce9e88b1a37a2ae95227d84abec377ccd3c5cfeb141ed4"
-dependencies = [
- "rustix",
- "windows-sys 0.59.0",
-]
-
 [[package]]
 name = "futures"
 version = "0.3.32"
@@ -857,7 +279,7 @@ checksum = "e835b70203e41293343137df5c0664546da5745f82ec9b84d40be8336958447b"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -889,25 +311,6 @@ dependencies = [
  "slab",
 ]
 
-[[package]]
-name = "fxhash"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c31b6d751ae2c7f11320402d34e41349dd1016f8d5d45e48c4312bc8625af50c"
-dependencies = [
- "byteorder",
-]
-
-[[package]]
-name = "generic-array"
-version = "0.14.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
-dependencies = [
- "typenum",
- "version_check",
-]
-
 [[package]]
 name = "getrandom"
 version = "0.2.17"
@@ -921,20 +324,6 @@ dependencies = [
  "wasm-bindgen",
 ]
 
-[[package]]
-name = "getrandom"
-version = "0.3.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "899def5c37c4fd7b2664648c28120ecec138e4d395b459e5ca34f9cce2dd77fd"
-dependencies = [
- "cfg-if",
- "js-sys",
- "libc",
- "r-efi 5.3.0",
- "wasip2",
- "wasm-bindgen",
-]
-
 [[package]]
 name = "getrandom"
 version = "0.4.2"
@@ -942,60 +331,21 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0de51e6874e94e7bf76d726fc5d13ba782deca734ff60d5bb2fb2607c7406555"
 dependencies = [
  "cfg-if",
+ "js-sys",
  "libc",
- "r-efi 6.0.0",
+ "r-efi",
+ "rand_core",
  "wasip2",
  "wasip3",
+ "wasm-bindgen",
 ]
 
-[[package]]
-name = "h2"
-version = "0.4.14"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "171fefbc92fe4a4de27e0698d6a5b392d6a0e333506bc49133760b3bcf948733"
-dependencies = [
- "atomic-waker",
- "bytes",
- "fnv",
- "futures-core",
- "futures-sink",
- "http",
- "indexmap 2.14.0",
- "slab",
- "tokio",
- "tokio-util",
- "tracing",
-]
-
-[[package]]
-name = "hash32"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b0c35f58762feb77d74ebe43bdbc3210f09be9fe6742234d573bacc26ed92b67"
-dependencies = [
- "byteorder",
-]
-
-[[package]]
-name = "hashbrown"
-version = "0.12.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8a9ee70c43aaf417c914396645a0fa852624801b24ebb7ae78fe8272889ac888"
-
-[[package]]
-name = "hashbrown"
-version = "0.14.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e5274423e17b7c9fc20b6e7e208532f9b19825d82dfd615708b70edd83df41f1"
-
 [[package]]
 name = "hashbrown"
 version = "0.15.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9229cfe53dfd69f0609a49f65461bd93001ea1ef889cd5529dd176593f5338a1"
 dependencies = [
- "allocator-api2",
- "equivalent",
  "foldhash",
 ]
 
@@ -1005,44 +355,12 @@ version = "0.17.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ed5909b6e89a2db4456e54cd5f673791d7eca6732202bbf2a9cc504fe2f9b84a"
 
-[[package]]
-name = "heapless"
-version = "0.7.17"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cdc6457c0eb62c71aac4bc17216026d8410337c4126773b9c5daba343f17964f"
-dependencies = [
- "atomic-polyfill",
- "hash32",
- "rustc_version",
- "serde",
- "spin",
- "stable_deref_trait",
-]
-
 [[package]]
 name = "heck"
 version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 
-[[package]]
-name = "hermit-abi"
-version = "0.5.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fc0fef456e4baa96da950455cd02c081ca953b141298e41db3fc7e36b1da849c"
-
-[[package]]
-name = "hex"
-version = "0.4.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7f24254aa9a54b5c858eaee2f5bccdb46aaf0e486a595ed5fd8f86ba55232a70"
-
-[[package]]
-name = "htmlescape"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e9025058dae765dee5070ec375f591e2ba14638c63feff74f13805a72e523163"
-
 [[package]]
 name = "http"
 version = "1.4.0"
@@ -1082,12 +400,6 @@ version = "1.10.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6dbf3de79e51f3d586ab4cb9d5c3e2c14aa28ed23d180cf89b4df0454a69cc87"
 
-[[package]]
-name = "httpdate"
-version = "1.0.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "df3b46402a9d5adb4c86a0cf463f42e19994e3ee891101b1841f30a545cb49a9"
-
 [[package]]
 name = "hyper"
 version = "1.9.0"
@@ -1098,11 +410,9 @@ dependencies = [
  "bytes",
  "futures-channel",
  "futures-core",
- "h2",
  "http",
  "http-body",
  "httparse",
- "httpdate",
  "itoa",
  "pin-project-lite",
  "smallvec",
@@ -1126,35 +436,6 @@ dependencies = [
  "webpki-roots",
 ]
 
-[[package]]
-name = "hyper-timeout"
-version = "0.5.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2b90d566bffbce6a75bd8b09a05aa8c2cb1fabb6cb348f8840c9e4c90a0d83b0"
-dependencies = [
- "hyper",
- "hyper-util",
- "pin-project-lite",
- "tokio",
- "tower-service",
-]
-
-[[package]]
-name = "hyper-tls"
-version = "0.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "70206fc6890eaca9fde8a0bf71caa2ddfc9fe045ac9e5c70df101a7dbde866e0"
-dependencies = [
- "bytes",
- "http-body-util",
- "hyper",
- "hyper-util",
- "native-tls",
- "tokio",
- "tokio-native-tls",
- "tower-service",
-]
-
 [[package]]
 name = "hyper-util"
 version = "0.1.20"
@@ -1172,21 +453,10 @@ dependencies = [
  "libc",
  "percent-encoding",
  "pin-project-lite",
- "socket2 0.6.3",
- "system-configuration",
+ "socket2 0.5.10",
  "tokio",
  "tower-service",
  "tracing",
- "windows-registry",
-]
-
-[[package]]
-name = "hyperloglogplus"
-version = "0.4.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "621debdf94dcac33e50475fdd76d34d5ea9c0362a834b9db08c3024696c1fbe3"
-dependencies = [
- "serde",
 ]
 
 [[package]]
@@ -1328,16 +598,6 @@ dependencies = [
  "icu_properties",
 ]
 
-[[package]]
-name = "indexmap"
-version = "1.9.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bd070e393353796e801d209ad339e89596eb4c8d430d18ede6a1cced8fafbd99"
-dependencies = [
- "autocfg",
- "hashbrown 0.12.3",
-]
-
 [[package]]
 name = "indexmap"
 version = "2.14.0"
@@ -1350,52 +610,18 @@ dependencies = [
  "serde_core",
 ]
 
-[[package]]
-name = "instant"
-version = "0.1.13"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e0242819d153cba4b4b05a5a8f2a7e9bbf97b6055b2a002b395c96b5ff3c0222"
-dependencies = [
- "cfg-if",
-]
-
 [[package]]
 name = "ipnet"
 version = "2.12.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d98f6fed1fde3f8c21bc40a1abb88dd75e67924f9cffc3ef95607bad8017f8e2"
 
-[[package]]
-name = "is_terminal_polyfill"
-version = "1.70.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a6cb138bb79a146c1bd460005623e142ef0181e3d0219cb493e02f7d08a35695"
-
-[[package]]
-name = "itertools"
-version = "0.14.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2b192c782037fadd9cfa75548310488aabdbf3d2da73885b31bd0abd03351285"
-dependencies = [
- "either",
-]
-
 [[package]]
 name = "itoa"
 version = "1.0.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8f42a60cbdf9a97f5d2305f08a87dc4e09308d1276d28c869c684d7777685682"
 
-[[package]]
-name = "jobserver"
-version = "0.1.34"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9afb3de4395d6b3e67a780b6de64b51c978ecf11cb9a462c66be7d4ca9039d33"
-dependencies = [
- "getrandom 0.3.4",
- "libc",
-]
-
 [[package]]
 name = "js-sys"
 version = "0.3.97"
@@ -1420,24 +646,12 @@ version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "09edd9e8b54e49e587e4f6295a7d29c3ea94d469cb40ab8ca70b288248a81db2"
 
-[[package]]
-name = "levenshtein_automata"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0c2cdeb66e45e9f36bfad5bbdb4d2384e70936afbee843c6f6543f0c551ebb25"
-
 [[package]]
 name = "libc"
 version = "0.2.186"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66"
 
-[[package]]
-name = "libm"
-version = "0.2.16"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b6d2cec3eae94f9f509c767b45932f1ada8350c4bdb85af2fcab4a3c14807981"
-
 [[package]]
 name = "libredox"
 version = "0.1.16"
@@ -1447,57 +661,24 @@ dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "linux-raw-sys"
-version = "0.12.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32a66949e030da00e8c7d4434b251670a91556f4144941d37452769c25d58a53"
-
 [[package]]
 name = "litemap"
 version = "0.8.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "92daf443525c4cce67b150400bc2316076100ce0b3686209eb8cf3c31612e6f0"
 
-[[package]]
-name = "lock_api"
-version = "0.4.14"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "224399e74b87b5f3557511d98dff8b14089b3dadafcab6bb93eab67d3aace965"
-dependencies = [
- "scopeguard",
-]
-
 [[package]]
 name = "log"
 version = "0.4.29"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5e5032e24019045c762d3c0f28f5b6b8bbf38563a65908389bf7978758920897"
 
-[[package]]
-name = "lru"
-version = "0.12.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "234cf4f4a04dc1f57e24b96cc0cd600cf2af460d4161ac5ecdd0af8e1f3b2a38"
-dependencies = [
- "hashbrown 0.15.5",
-]
-
 [[package]]
 name = "lru-slab"
 version = "0.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "112b39cec0b298b6c1999fee3e31427f74f676e4cb9879ed1a121b43661a4154"
 
-[[package]]
-name = "lz4_flex"
-version = "0.11.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "373f5eceeeab7925e0c1098212f2fbc4d416adec9d35051a6ab251e824c1854a"
-dependencies = [
- "twox-hash",
-]
-
 [[package]]
 name = "matchers"
 version = "0.2.0"
@@ -1507,58 +688,12 @@ dependencies = [
  "regex-automata",
 ]
 
-[[package]]
-name = "matchit"
-version = "0.7.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0e7465ac9959cc2b1404e8e2367b43684a6d13790fe23056cc8c6c5a6b7bcb94"
-
-[[package]]
-name = "measure_time"
-version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "51c55d61e72fc3ab704396c5fa16f4c184db37978ae4e94ca8959693a235fc0e"
-dependencies = [
- "log",
-]
-
 [[package]]
 name = "memchr"
 version = "2.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f8ca58f447f06ed17d5fc4043ce1b10dd205e060fb3ce5b979b8ed8e59ff3f79"
 
-[[package]]
-name = "memmap2"
-version = "0.9.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "714098028fe011992e1c3962653c96b2d578c4b4bce9036e15ff220319b1e0e3"
-dependencies = [
- "libc",
-]
-
-[[package]]
-name = "mime"
-version = "0.3.17"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6877bb514081ee2a7ff5ef9de3281f14a4dd4bceac4c09388074a6b5df8a139a"
-
-[[package]]
-name = "minimal-lexical"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a"
-
-[[package]]
-name = "miniz_oxide"
-version = "0.8.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fa76a2c86f704bdb222d66965fb3d63269ce38518b83cb0575fca855ebb6316"
-dependencies = [
- "adler2",
- "simd-adler32",
-]
-
 [[package]]
 name = "mio"
 version = "1.2.0"
@@ -1570,54 +705,15 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
-[[package]]
-name = "murmurhash32"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2195bf6aa996a481483b29d62a7663eed3fe39600c460e323f8ff41e90bdd89b"
-
-[[package]]
-name = "native-tls"
-version = "0.2.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "465500e14ea162429d264d44189adc38b199b62b1c21eea9f69e4b73cb03bbf2"
-dependencies = [
- "libc",
- "log",
- "openssl",
- "openssl-probe",
- "openssl-sys",
- "schannel",
- "security-framework",
- "security-framework-sys",
- "tempfile",
-]
-
-[[package]]
-name = "nom"
-version = "7.1.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d273983c5a657a70a3e8f2a01329822f3b8c8172b73826411a55751e404a0a4a"
-dependencies = [
- "memchr",
- "minimal-lexical",
-]
-
 [[package]]
 name = "nu-ansi-term"
 version = "0.50.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
 dependencies = [
- "windows-sys 0.61.2",
+ "windows-sys 0.59.0",
 ]
 
-[[package]]
-name = "num-conv"
-version = "0.2.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "521739c6d2bac4aa25192232afe6841231376b2b26d4d9fae5ecf8ca5772e441"
-
 [[package]]
 name = "num-traits"
 version = "0.2.19"
@@ -1625,17 +721,6 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
 dependencies = [
  "autocfg",
- "libm",
-]
-
-[[package]]
-name = "num_cpus"
-version = "1.17.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91df4bbde75afed763b708b7eee1e8e7651e02d97f6d5dd763e89367e957b23b"
-dependencies = [
- "hermit-abi",
- "libc",
 ]
 
 [[package]]
@@ -1644,61 +729,6 @@ version = "1.21.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
 
-[[package]]
-name = "once_cell_polyfill"
-version = "1.70.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "384b8ab6d37215f3c5301a95a4accb5d64aa607f1fcb26a11b5303878451b4fe"
-
-[[package]]
-name = "oneshot"
-version = "0.1.13"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "269bca4c2591a28585d6bf10d9ed0332b7d76900a1b02bec41bdc3a2cdcda107"
-
-[[package]]
-name = "openssl"
-version = "0.10.80"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a45fa2aa886c42762255da344f0a0d313e254066c46aad76f300c3d3da62d967"
-dependencies = [
- "bitflags 2.11.1",
- "cfg-if",
- "foreign-types",
- "libc",
- "openssl-macros",
- "openssl-sys",
-]
-
-[[package]]
-name = "openssl-macros"
-version = "0.1.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a948666b637a0f465e8564c73e89d4dde00d72d4d473cc972f390fc3dcee7d9c"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "openssl-probe"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7c87def4c32ab89d880effc9e097653c8da5d6ef28e6b539d313baaacfbafcbe"
-
-[[package]]
-name = "openssl-sys"
-version = "0.9.116"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f28a22dc7140cda5f096e5e7724a6962ca81a7f8bfd2979f9b18c11af56318c4"
-dependencies = [
- "cc",
- "libc",
- "pkg-config",
- "vcpkg",
-]
-
 [[package]]
 name = "opentelemetry"
 version = "0.32.0"
@@ -1709,7 +739,7 @@ dependencies = [
  "futures-sink",
  "js-sys",
  "pin-project-lite",
- "thiserror 2.0.18",
+ "thiserror",
  "tracing",
 ]
 
@@ -1719,63 +749,6 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "04744f49eae99ab78e0d5c0b603ab218f515ea8cfe5a456d7629ad883a3b6e7d"
 
-[[package]]
-name = "ownedbytes"
-version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2fbd56f7631767e61784dc43f8580f403f4475bd4aaa4da003e6295e1bab4a7e"
-dependencies = [
- "stable_deref_trait",
-]
-
-[[package]]
-name = "parking_lot"
-version = "0.11.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7d17b78036a60663b797adeaee46f5c9dfebb86948d1255007a1d6be0271ff99"
-dependencies = [
- "instant",
- "lock_api",
- "parking_lot_core 0.8.6",
-]
-
-[[package]]
-name = "parking_lot"
-version = "0.12.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "93857453250e3077bd71ff98b6a65ea6621a19bb0f559a85248955ac12c45a1a"
-dependencies = [
- "lock_api",
- "parking_lot_core 0.9.12",
-]
-
-[[package]]
-name = "parking_lot_core"
-version = "0.8.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "60a2cfe6f0ad2bfc16aefa463b497d5c7a5ecd44a23efa72aa342d90177356dc"
-dependencies = [
- "cfg-if",
- "instant",
- "libc",
- "redox_syscall 0.2.16",
- "smallvec",
- "winapi",
-]
-
-[[package]]
-name = "parking_lot_core"
-version = "0.9.12"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2621685985a2ebf1c516881c026032ac7deafcda1a2c9b7850dc81e3dfcb64c1"
-dependencies = [
- "cfg-if",
- "libc",
- "redox_syscall 0.5.18",
- "smallvec",
- "windows-link",
-]
-
 [[package]]
 name = "pastey"
 version = "0.2.2"
@@ -1788,51 +761,12 @@ version = "2.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9b4f627cb1b25917193a259e49bdad08f671f8d9708acfd5fe0a8c1455d87220"
 
-[[package]]
-name = "pin-project"
-version = "1.1.13"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2466b2336ed02bcdca6b294417127b90ec92038d1d5c4fbeac971a922e0e0924"
-dependencies = [
- "pin-project-internal",
-]
-
-[[package]]
-name = "pin-project-internal"
-version = "1.1.13"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c96395f0a926bc13b1c17622aaddda1ecb55d49c8f1bf9777e4d877800a43f8b"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
 [[package]]
 name = "pin-project-lite"
 version = "0.2.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a89322df9ebe1c1578d689c92318e070967d1042b512afbe49518723f4e6d5cd"
 
-[[package]]
-name = "pkg-config"
-version = "0.3.33"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "19f132c84eca552bf34cab8ec81f1c1dcc229b811638f9d283dceabe58c5569e"
-
-[[package]]
-name = "postcard"
-version = "1.1.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6764c3b5dd454e283a30e6dfe78e9b31096d9e32036b5d1eaac7a6119ccb9a24"
-dependencies = [
- "cobs",
- "embedded-io 0.4.0",
- "embedded-io 0.6.1",
- "heapless",
- "serde",
-]
-
 [[package]]
 name = "potential_utf"
 version = "0.1.5"
@@ -1842,21 +776,6 @@ dependencies = [
  "zerovec",
 ]
 
-[[package]]
-name = "powerfmt"
-version = "0.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
-
-[[package]]
-name = "ppv-lite86"
-version = "0.2.21"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "85eae3c4ed2f50dcfe72643da4befc30deadb458a9b590d720cde2f2b1e97da9"
-dependencies = [
- "zerocopy",
-]
-
 [[package]]
 name = "prettyplease"
 version = "0.2.37"
@@ -1864,7 +783,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "479ca8adacdd7ce8f1fb39ce9ecccbfe93a3f1344b3d0d97f20bc0196208f62b"
 dependencies = [
  "proc-macro2",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -1876,60 +795,6 @@ dependencies = [
  "unicode-ident",
 ]
 
-[[package]]
-name = "prost"
-version = "0.13.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2796faa41db3ec313a31f7624d9286acf277b52de526150b7e69f3debf891ee5"
-dependencies = [
- "bytes",
- "prost-derive",
-]
-
-[[package]]
-name = "prost-derive"
-version = "0.13.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8a56d757972c98b346a9b766e3f02746cde6dd1cd1d1d563472929fdd74bec4d"
-dependencies = [
- "anyhow",
- "itertools",
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "prost-types"
-version = "0.13.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "52c2c1bf36ddb1a1c396b3601a3cec27c2462e45f07c386894ec3ccf5332bd16"
-dependencies = [
- "prost",
-]
-
-[[package]]
-name = "qdrant-client"
-version = "1.18.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "82cef4e669bcf9c07471463adab5ee080dd9bc9381f3652ea4981f6030b2c309"
-dependencies = [
- "anyhow",
- "derive_builder",
- "futures",
- "futures-util",
- "parking_lot 0.12.5",
- "prost",
- "prost-types",
- "reqwest",
- "semver",
- "serde",
- "serde_json",
- "thiserror 1.0.69",
- "tokio",
- "tonic",
-]
-
 [[package]]
 name = "quinn"
 version = "0.11.9"
@@ -1943,8 +808,8 @@ dependencies = [
  "quinn-udp",
  "rustc-hash",
  "rustls",
- "socket2 0.6.3",
- "thiserror 2.0.18",
+ "socket2 0.5.10",
+ "thiserror",
  "tokio",
  "tracing",
  "web-time",
@@ -1952,20 +817,21 @@ dependencies = [
 
 [[package]]
 name = "quinn-proto"
-version = "0.11.14"
+version = "0.11.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "434b42fec591c96ef50e21e886936e66d3cc3f737104fdb9b737c40ffb94c098"
+checksum = "04759210543be93709136e28212294a659ef5001836ff4eab4d663e4529bba83"
 dependencies = [
  "bytes",
- "getrandom 0.3.4",
+ "getrandom 0.4.2",
  "lru-slab",
- "rand 0.9.4",
+ "rand",
+ "rand_pcg",
  "ring",
  "rustc-hash",
  "rustls",
  "rustls-pki-types",
  "slab",
- "thiserror 2.0.18",
+ "thiserror",
  "tinyvec",
  "tracing",
  "web-time",
@@ -1980,7 +846,7 @@ dependencies = [
  "cfg_aliases",
  "libc",
  "once_cell",
- "socket2 0.6.3",
+ "socket2 0.5.10",
  "tracing",
  "windows-sys 0.60.2",
 ]
@@ -1994,12 +860,6 @@ dependencies = [
  "proc-macro2",
 ]
 
-[[package]]
-name = "r-efi"
-version = "5.3.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "69cdb34c158ceb288df11e18b4bd39de994f6657d83847bdffdbd7f346754b0f"
-
 [[package]]
 name = "r-efi"
 version = "6.0.0"
@@ -2007,138 +867,40 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f8dcc9c7d52a811697d2151c701e0d08956f92b0e24136cf4cf27b57a6a0d9bf"
 
 [[package]]
-name = "rand"
-version = "0.8.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5ca0ecfa931c29007047d1bc58e623ab12e5590e8c7cc53200d5202b69266d8a"
-dependencies = [
- "libc",
- "rand_chacha 0.3.1",
- "rand_core 0.6.4",
-]
-
-[[package]]
-name = "rand"
-version = "0.9.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "44c5af06bb1b7d3216d91932aed5265164bf384dc89cd6ba05cf59a35f5f76ea"
-dependencies = [
- "rand_chacha 0.9.0",
- "rand_core 0.9.5",
-]
-
-[[package]]
-name = "rand_chacha"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e6c10a63a0fa32252be49d21e7709d4d4baf8d231c2dbce1eaa8141b9b127d88"
-dependencies = [
- "ppv-lite86",
- "rand_core 0.6.4",
-]
-
-[[package]]
-name = "rand_chacha"
-version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d3022b5f1df60f26e1ffddd6c66e8aa15de382ae63b3a0c1bfc0e4d3e3f325cb"
-dependencies = [
- "ppv-lite86",
- "rand_core 0.9.5",
-]
-
-[[package]]
-name = "rand_core"
-version = "0.6.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ec0be4795e2f6a28069bec0b5ff3e2ac9bafc99e6a9a7dc3547996c5c816922c"
-dependencies = [
- "getrandom 0.2.17",
-]
-
-[[package]]
-name = "rand_core"
-version = "0.9.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "76afc826de14238e6e8c374ddcc1fa19e374fd8dd986b0d2af0d02377261d83c"
-dependencies = [
- "getrandom 0.3.4",
-]
-
-[[package]]
-name = "rand_distr"
-version = "0.4.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32cb0b9bc82b0a0876c2dd994a7e7a2683d3e7390ca40e6886785ef0c7e3ee31"
-dependencies = [
- "num-traits",
- "rand 0.8.6",
-]
-
-[[package]]
-name = "rayon"
-version = "1.12.0"
+name = "rand"
+version = "0.10.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb39b166781f92d482534ef4b4b1b2568f42613b53e5b6c160e24cfbfa30926d"
+checksum = "c7f5fa3a058cd35567ef9bfa5e75732bee0f9e4c55fa90477bef2dfcdbc4be80"
 dependencies = [
- "either",
- "rayon-core",
+ "chacha20",
+ "getrandom 0.4.2",
+ "rand_core",
 ]
 
 [[package]]
-name = "rayon-core"
-version = "1.13.0"
+name = "rand_core"
+version = "0.10.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "22e18b0f0062d30d4230b2e85ff77fdfe4326feb054b9783a3460d8435c8ab91"
-dependencies = [
- "crossbeam-deque",
- "crossbeam-utils",
-]
+checksum = "63b8176103e19a2643978565ca18b50549f6101881c443590420e4dc998a3c69"
 
 [[package]]
-name = "reasonkit-mem"
-version = "0.1.7"
+name = "rand_pcg"
+version = "0.10.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f5baf8557e8b9a73636cea0cd6f17d7d06e5d9762d200431b5f7b5f3bae7ab68"
+checksum = "caa0f4137e1c0a72f4c651489402276c8e8e1cf081f3b0ba156d2cbeef09e86a"
 dependencies = [
- "anyhow",
- "async-trait",
- "chrono",
- "clap",
- "crc32fast",
- "crossbeam-utils",
- "dashmap",
- "dirs",
- "hex",
- "lz4_flex",
- "num_cpus",
- "postcard",
- "qdrant-client",
- "rayon",
- "reqwest",
- "rmp-serde",
- "serde",
- "serde_bytes",
- "serde_json",
- "sha2",
- "sled",
- "tantivy",
- "thiserror 1.0.69",
- "tokio",
- "tracing",
- "uuid",
+ "rand_core",
 ]
 
 [[package]]
 name = "reasonkit-think-mcp"
-version = "0.1.3"
+version = "0.2.0"
 dependencies = [
  "anyhow",
  "chrono",
  "directories",
  "once_cell",
  "opentelemetry",
- "reasonkit-mem",
  "regex",
  "reqwest",
  "rmcp",
@@ -2151,35 +913,6 @@ dependencies = [
  "uuid",
 ]
 
-[[package]]
-name = "redox_syscall"
-version = "0.2.16"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb5a58c1855b4b6819d59012155603f0b22ad30cad752600aadfcb695265519a"
-dependencies = [
- "bitflags 1.3.2",
-]
-
-[[package]]
-name = "redox_syscall"
-version = "0.5.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ed2bf2547551a7053d6fdfafda3f938979645c44812fbfcda098faae3f1a362d"
-dependencies = [
- "bitflags 2.11.1",
-]
-
-[[package]]
-name = "redox_users"
-version = "0.4.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ba009ff324d1fc1b900bd1fdb31564febe58a8ccc8a6fdbb93b543d33b13ca43"
-dependencies = [
- "getrandom 0.2.17",
- "libredox",
- "thiserror 1.0.69",
-]
-
 [[package]]
 name = "redox_users"
 version = "0.5.2"
@@ -2188,7 +921,7 @@ checksum = "a4e608c6638b9c18977b00b475ac1f28d14e84b27d8d42f70e0bf1e3dec127ac"
 dependencies = [
  "getrandom 0.2.17",
  "libredox",
- "thiserror 2.0.18",
+ "thiserror",
 ]
 
 [[package]]
@@ -2208,7 +941,7 @@ checksum = "b7186006dcb21920990093f30e3dea63b7d6e977bf1256be20c3563a5db070da"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2248,22 +981,17 @@ checksum = "eddd3ca559203180a307f12d114c268abf583f59b03cb906fd0b3ff8646c1147"
 dependencies = [
  "base64",
  "bytes",
- "encoding_rs",
  "futures-channel",
  "futures-core",
  "futures-util",
- "h2",
  "http",
  "http-body",
  "http-body-util",
  "hyper",
  "hyper-rustls",
- "hyper-tls",
  "hyper-util",
  "js-sys",
  "log",
- "mime",
- "native-tls",
  "percent-encoding",
  "pin-project-lite",
  "quinn",
@@ -2274,16 +1002,13 @@ dependencies = [
  "serde_urlencoded",
  "sync_wrapper",
  "tokio",
- "tokio-native-tls",
  "tokio-rustls",
- "tokio-util",
- "tower 0.5.3",
+ "tower",
  "tower-http",
  "tower-service",
  "url",
  "wasm-bindgen",
  "wasm-bindgen-futures",
- "wasm-streams",
  "web-sys",
  "webpki-roots",
 ]
@@ -2304,65 +1029,37 @@ dependencies = [
 
 [[package]]
 name = "rmcp"
-version = "1.6.0"
+version = "3.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e12ca9067b5ebfbd5b3fcdc4acfceb81aa7d5ab2a879dff7cb75d22434276aad"
+checksum = "1a15bc53261a9dc37e105df006e4656c598379a8f9581f8950debb130f27a7cf"
 dependencies = [
- "async-trait",
  "chrono",
  "futures",
+ "indexmap",
  "pastey",
  "pin-project-lite",
  "rmcp-macros",
  "schemars",
  "serde",
  "serde_json",
- "thiserror 2.0.18",
+ "thiserror",
  "tokio",
  "tokio-util",
  "tracing",
+ "uuid",
 ]
 
 [[package]]
 name = "rmcp-macros"
-version = "1.6.0"
+version = "3.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7caa6743cc0888e433105fe1bc551a7f607940b126a37bc97b478e86064627eb"
+checksum = "a85d45508e9b4ba024fe996c2638799635d75b6dd0ba8f32ccf08f8026f0c780"
 dependencies = [
- "darling 0.23.0",
+ "darling",
  "proc-macro2",
  "quote",
  "serde_json",
- "syn",
-]
-
-[[package]]
-name = "rmp"
-version = "0.8.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4ba8be72d372b2c9b35542551678538b562e7cf86c3315773cae48dfbfe7790c"
-dependencies = [
- "num-traits",
-]
-
-[[package]]
-name = "rmp-serde"
-version = "1.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "72f81bee8c8ef9b577d1681a70ebbc962c232461e397b22c208c43c04b67a155"
-dependencies = [
- "rmp",
- "serde",
-]
-
-[[package]]
-name = "rust-stemmers"
-version = "1.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e46a2036019fdb888131db7a4c847a1063a7493f971ed94ea82c67eada63ca54"
-dependencies = [
- "serde",
- "serde_derive",
+ "syn 3.0.3",
 ]
 
 [[package]]
@@ -2371,35 +1068,12 @@ version = "2.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "94300abf3f1ae2e2b8ffb7b58043de3d399c73fa6f4b73826402a5c457614dbe"
 
-[[package]]
-name = "rustc_version"
-version = "0.4.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cfcb3a22ef46e85b45de6ee7e79d063319ebb6594faafcf1c225ea92ab6e9b92"
-dependencies = [
- "semver",
-]
-
-[[package]]
-name = "rustix"
-version = "1.1.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b6fe4565b9518b83ef4f91bb47ce29620ca828bd32cb7e408f0062e9930ba190"
-dependencies = [
- "bitflags 2.11.1",
- "errno",
- "libc",
- "linux-raw-sys",
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "rustls"
 version = "0.23.40"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ef86cd5876211988985292b91c96a8f2d298df24e75989a43a3c73f2d4d8168b"
 dependencies = [
- "log",
  "once_cell",
  "ring",
  "rustls-pki-types",
@@ -2408,27 +1082,6 @@ dependencies = [
  "zeroize",
 ]
 
-[[package]]
-name = "rustls-native-certs"
-version = "0.8.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "612460d5f7bea540c490b2b6395d8e34a953e52b491accd6c86c8164c5932a63"
-dependencies = [
- "openssl-probe",
- "rustls-pki-types",
- "schannel",
- "security-framework",
-]
-
-[[package]]
-name = "rustls-pemfile"
-version = "2.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dce314e5fee3f39953d46bb63bb8a46d40c2f8fb7cc5a3b6cab2bde9721d6e50"
-dependencies = [
- "rustls-pki-types",
-]
-
 [[package]]
 name = "rustls-pki-types"
 version = "1.14.1"
@@ -2462,15 +1115,6 @@ version = "1.0.23"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f"
 
-[[package]]
-name = "schannel"
-version = "0.1.29"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91c1b7e4904c873ef0710c1f407dde2e6287de2bebc1bbbf7d430bb7cbffd939"
-dependencies = [
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "schemars"
 version = "1.2.1"
@@ -2494,36 +1138,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "serde_derive_internals",
- "syn",
-]
-
-[[package]]
-name = "scopeguard"
-version = "1.2.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
-
-[[package]]
-name = "security-framework"
-version = "3.7.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b7f4bc775c73d9a02cde8bf7b2ec4c9d12743edf609006c7facc23998404cd1d"
-dependencies = [
- "bitflags 2.11.1",
- "core-foundation 0.10.1",
- "core-foundation-sys",
- "libc",
- "security-framework-sys",
-]
-
-[[package]]
-name = "security-framework-sys"
-version = "2.17.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ce2691df843ecc5d231c0b14ece2acc3efb62c0a398c7e1d875f3983ce020e3"
-dependencies = [
- "core-foundation-sys",
- "libc",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2542,16 +1157,6 @@ dependencies = [
  "serde_derive",
 ]
 
-[[package]]
-name = "serde_bytes"
-version = "0.11.19"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a5d440709e79d88e51ac01c4b72fc6cb7314017bb7da9eeff678aa94c10e3ea8"
-dependencies = [
- "serde",
- "serde_core",
-]
-
 [[package]]
 name = "serde_core"
 version = "1.0.228"
@@ -2569,7 +1174,7 @@ checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2580,7 +1185,7 @@ checksum = "18d26a20a969b9e3fdf2fc2d9f21eda6c40e2de84c9408bb5d3b05d499aae711"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2608,17 +1213,6 @@ dependencies = [
  "serde",
 ]
 
-[[package]]
-name = "sha2"
-version = "0.10.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a7507d819769d01a365ab707794a4084392c824f54a7a6a7862f8c3d0892b283"
-dependencies = [
- "cfg-if",
- "cpufeatures",
- "digest",
-]
-
 [[package]]
 name = "sharded-slab"
 version = "0.1.7"
@@ -2634,324 +1228,96 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"
 
-[[package]]
-name = "signal-hook-registry"
-version = "1.4.8"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c4db69cba1110affc0e9f7bcd48bbf87b3f4fc7c61fc9155afd4c469eb3d6c1b"
-dependencies = [
- "errno",
- "libc",
-]
-
-[[package]]
-name = "simd-adler32"
-version = "0.3.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "703d5c7ef118737c72f1af64ad2f6f8c5e1921f818cdcb97b8fe6fc69bf66214"
-
-[[package]]
-name = "sketches-ddsketch"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0c6f73aeb92d671e0cc4dca167e59b2deb6387c375391bc99ee743f326994a2b"
-dependencies = [
- "serde",
-]
-
 [[package]]
 name = "slab"
 version = "0.4.12"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0c790de23124f9ab44544d7ac05d60440adc586479ce501c1d6d7da3cd8c9cf5"
-
-[[package]]
-name = "sled"
-version = "0.34.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7f96b4737c2ce5987354855aed3797279def4ebf734436c6aa4552cf8e169935"
-dependencies = [
- "crc32fast",
- "crossbeam-epoch",
- "crossbeam-utils",
- "fs2",
- "fxhash",
- "libc",
- "log",
- "parking_lot 0.11.2",
-]
-
-[[package]]
-name = "smallvec"
-version = "1.15.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "67b1b7a3b5fe4f1376887184045fcf45c69e92af734b7aaddc05fb777b6fbd03"
-
-[[package]]
-name = "socket2"
-version = "0.5.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e22376abed350d73dd1cd119b57ffccad95b4e585a7cda43e286245ce23c0678"
-dependencies = [
- "libc",
- "windows-sys 0.52.0",
-]
-
-[[package]]
-name = "socket2"
-version = "0.6.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3a766e1110788c36f4fa1c2b71b387a7815aa65f88ce0229841826633d93723e"
-dependencies = [
- "libc",
- "windows-sys 0.61.2",
-]
-
-[[package]]
-name = "spin"
-version = "0.9.8"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6980e8d7511241f8acf4aebddbb1ff938df5eebe98691418c4468d0b72a96a67"
-dependencies = [
- "lock_api",
-]
-
-[[package]]
-name = "stable_deref_trait"
-version = "1.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ce2be8dc25455e1f91df71bfa12ad37d7af1092ae736f3a6cd0e37bc7810596"
-
-[[package]]
-name = "strsim"
-version = "0.11.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7da8b5736845d9f2fcb837ea5d9e2628564b3b043a70948a3f0b778838c5fb4f"
-
-[[package]]
-name = "subtle"
-version = "2.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"
-
-[[package]]
-name = "syn"
-version = "2.0.117"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e665b8803e7b1d2a727f4023456bbbbe74da67099c585258af0ad9c5013b9b99"
-dependencies = [
- "proc-macro2",
- "quote",
- "unicode-ident",
-]
-
-[[package]]
-name = "sync_wrapper"
-version = "1.0.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0bf256ce5efdfa370213c1dabab5935a12e49f2c58d15e9eac2870d3b4f27263"
-dependencies = [
- "futures-core",
-]
-
-[[package]]
-name = "synstructure"
-version = "0.13.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "728a70f3dbaf5bab7f0c4b1ac8d7ae5ea60a4b5549c8a5914361c99147a709d2"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
-[[package]]
-name = "system-configuration"
-version = "0.7.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a13f3d0daba03132c0aa9767f98351b3488edc2c100cda2d2ec2b04f3d8d3c8b"
-dependencies = [
- "bitflags 2.11.1",
- "core-foundation 0.9.4",
- "system-configuration-sys",
-]
-
-[[package]]
-name = "system-configuration-sys"
-version = "0.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8e1d1b10ced5ca923a1fcb8d03e96b8d3268065d724548c0211415ff6ac6bac4"
-dependencies = [
- "core-foundation-sys",
- "libc",
-]
-
-[[package]]
-name = "tantivy"
-version = "0.25.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "502915c7381c5cb2d2781503962610cb880ad8f1a0ca95df1bae645d5ebf2545"
-dependencies = [
- "aho-corasick",
- "arc-swap",
- "base64",
- "bitpacking",
- "bon",
- "byteorder",
- "census",
- "crc32fast",
- "crossbeam-channel",
- "downcast-rs",
- "fastdivide",
- "fnv",
- "fs4",
- "htmlescape",
- "hyperloglogplus",
- "itertools",
- "levenshtein_automata",
- "log",
- "lru",
- "lz4_flex",
- "measure_time",
- "memmap2",
- "once_cell",
- "oneshot",
- "rayon",
- "regex",
- "rust-stemmers",
- "rustc-hash",
- "serde",
- "serde_json",
- "sketches-ddsketch",
- "smallvec",
- "tantivy-bitpacker",
- "tantivy-columnar",
- "tantivy-common",
- "tantivy-fst",
- "tantivy-query-grammar",
- "tantivy-stacker",
- "tantivy-tokenizer-api",
- "tempfile",
- "thiserror 2.0.18",
- "time",
- "uuid",
- "winapi",
-]
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c790de23124f9ab44544d7ac05d60440adc586479ce501c1d6d7da3cd8c9cf5"
 
 [[package]]
-name = "tantivy-bitpacker"
-version = "0.9.0"
+name = "smallvec"
+version = "1.15.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c3b04eed5108d8283607da6710fe17a7663523440eaf7ea5a1a440d19a1448b6"
-dependencies = [
- "bitpacking",
-]
+checksum = "67b1b7a3b5fe4f1376887184045fcf45c69e92af734b7aaddc05fb777b6fbd03"
 
 [[package]]
-name = "tantivy-columnar"
-version = "0.6.0"
+name = "socket2"
+version = "0.5.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8b628488ae936c83e92b5c4056833054ca56f76c0e616aee8339e24ac89119cd"
+checksum = "e22376abed350d73dd1cd119b57ffccad95b4e585a7cda43e286245ce23c0678"
 dependencies = [
- "downcast-rs",
- "fastdivide",
- "itertools",
- "serde",
- "tantivy-bitpacker",
- "tantivy-common",
- "tantivy-sstable",
- "tantivy-stacker",
+ "libc",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
-name = "tantivy-common"
-version = "0.10.0"
+name = "socket2"
+version = "0.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f880aa7cab0c063a47b62596d10991cdd0b6e0e0575d9c5eeb298b307a25de55"
+checksum = "3a766e1110788c36f4fa1c2b71b387a7815aa65f88ce0229841826633d93723e"
 dependencies = [
- "async-trait",
- "byteorder",
- "ownedbytes",
- "serde",
- "time",
+ "libc",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
-name = "tantivy-fst"
-version = "0.5.0"
+name = "stable_deref_trait"
+version = "1.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d60769b80ad7953d8a7b2c70cdfe722bbcdcac6bccc8ac934c40c034d866fc18"
-dependencies = [
- "byteorder",
- "regex-syntax",
- "utf8-ranges",
-]
+checksum = "6ce2be8dc25455e1f91df71bfa12ad37d7af1092ae736f3a6cd0e37bc7810596"
 
 [[package]]
-name = "tantivy-query-grammar"
-version = "0.25.0"
+name = "strsim"
+version = "0.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "768fccdc84d60d86235d42d7e4c33acf43c418258ff5952abf07bd7837fcd26b"
-dependencies = [
- "nom",
- "serde",
- "serde_json",
-]
+checksum = "7da8b5736845d9f2fcb837ea5d9e2628564b3b043a70948a3f0b778838c5fb4f"
 
 [[package]]
-name = "tantivy-sstable"
-version = "0.6.0"
+name = "subtle"
+version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f8292095d1a8a2c2b36380ec455f910ab52dde516af36321af332c93f20ab7d5"
-dependencies = [
- "futures-util",
- "itertools",
- "tantivy-bitpacker",
- "tantivy-common",
- "tantivy-fst",
- "zstd",
-]
+checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"
 
 [[package]]
-name = "tantivy-stacker"
-version = "0.6.0"
+name = "syn"
+version = "2.0.117"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "23d38a379411169f0b3002c9cba61cdfe315f757e9d4f239c00c282497a0749d"
+checksum = "e665b8803e7b1d2a727f4023456bbbbe74da67099c585258af0ad9c5013b9b99"
 dependencies = [
- "murmurhash32",
- "rand_distr",
- "tantivy-common",
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
 ]
 
 [[package]]
-name = "tantivy-tokenizer-api"
-version = "0.6.0"
+name = "syn"
+version = "3.0.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "23024f6aeb25ceb1a0e27740c84bdb0fae52626737b7e9a9de6ad5aa25c7b038"
+checksum = "53e9bae58849f64dfa4f5d5ae372c8341f7305f82a3868709269343628b659a3"
 dependencies = [
- "serde",
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
 ]
 
 [[package]]
-name = "tempfile"
-version = "3.27.0"
+name = "sync_wrapper"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd"
+checksum = "0bf256ce5efdfa370213c1dabab5935a12e49f2c58d15e9eac2870d3b4f27263"
 dependencies = [
- "fastrand",
- "getrandom 0.4.2",
- "once_cell",
- "rustix",
- "windows-sys 0.61.2",
+ "futures-core",
 ]
 
 [[package]]
-name = "thiserror"
-version = "1.0.69"
+name = "synstructure"
+version = "0.13.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b6aaf5339b578ea85b50e080feb250a3e8ae8cfcdff9a461c9ec2904bc923f52"
+checksum = "728a70f3dbaf5bab7f0c4b1ac8d7ae5ea60a4b5549c8a5914361c99147a709d2"
 dependencies = [
- "thiserror-impl 1.0.69",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2960,18 +1326,7 @@ version = "2.0.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4288b5bcbc7920c07a1149a35cf9590a2aa808e0bc1eafaade0b80947865fbc4"
 dependencies = [
- "thiserror-impl 2.0.18",
-]
-
-[[package]]
-name = "thiserror-impl"
-version = "1.0.69"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4fee6c4efc90059e10f81e6d42c60a18f76588c3d74cb83a0b242a2b6c7504c1"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
+ "thiserror-impl",
 ]
 
 [[package]]
@@ -2982,7 +1337,7 @@ checksum = "ebc4ee7f67670e9b64d05fa4253e753e016c6c95ff35b89b7941d6b856dec1d5"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -2994,37 +1349,6 @@ dependencies = [
  "cfg-if",
 ]
 
-[[package]]
-name = "time"
-version = "0.3.47"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "743bd48c283afc0388f9b8827b976905fb217ad9e647fae3a379a9283c4def2c"
-dependencies = [
- "deranged",
- "itoa",
- "num-conv",
- "powerfmt",
- "serde_core",
- "time-core",
- "time-macros",
-]
-
-[[package]]
-name = "time-core"
-version = "0.1.8"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7694e1cfe791f8d31026952abf09c69ca6f6fa4e1a1229e18988f06a04a12dca"
-
-[[package]]
-name = "time-macros"
-version = "0.2.27"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2e70e4c5a0e0a8a4823ad65dfe1a6930e4f4d756dcd9dd7939022b5e8c501215"
-dependencies = [
- "num-conv",
- "time-core",
-]
-
 [[package]]
 name = "tinystr"
 version = "0.8.3"
@@ -3059,9 +1383,7 @@ dependencies = [
  "bytes",
  "libc",
  "mio",
- "parking_lot 0.12.5",
  "pin-project-lite",
- "signal-hook-registry",
  "socket2 0.6.3",
  "tokio-macros",
  "windows-sys 0.61.2",
@@ -3075,17 +1397,7 @@ checksum = "385a6cb71ab9ab790c5fe8d67f1645e6c450a7ce006a33de03daa956cf70a496"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
-]
-
-[[package]]
-name = "tokio-native-tls"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bbae76ab933c85776efabc971569dd6119c580d8f5d448769dec1764bf796ef2"
-dependencies = [
- "native-tls",
- "tokio",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -3098,17 +1410,6 @@ dependencies = [
  "tokio",
 ]
 
-[[package]]
-name = "tokio-stream"
-version = "0.1.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32da49809aab5c3bc678af03902d4ccddea2a87d028d86392a4b1560c6906c70"
-dependencies = [
- "futures-core",
- "pin-project-lite",
- "tokio",
-]
-
 [[package]]
 name = "tokio-util"
 version = "0.7.18"
@@ -3122,60 +1423,6 @@ dependencies = [
  "tokio",
 ]
 
-[[package]]
-name = "tonic"
-version = "0.12.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "877c5b330756d856ffcc4553ab34a5684481ade925ecc54bcd1bf02b1d0d4d52"
-dependencies = [
- "async-stream",
- "async-trait",
- "axum",
- "base64",
- "bytes",
- "flate2",
- "h2",
- "http",
- "http-body",
- "http-body-util",
- "hyper",
- "hyper-timeout",
- "hyper-util",
- "percent-encoding",
- "pin-project",
- "prost",
- "rustls-native-certs",
- "rustls-pemfile",
- "socket2 0.5.10",
- "tokio",
- "tokio-rustls",
- "tokio-stream",
- "tower 0.4.13",
- "tower-layer",
- "tower-service",
- "tracing",
-]
-
-[[package]]
-name = "tower"
-version = "0.4.13"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b8fa9be0de6cf49e536ce1851f987bd21a43b771b09473c3549a6c853db37c1c"
-dependencies = [
- "futures-core",
- "futures-util",
- "indexmap 1.9.3",
- "pin-project",
- "pin-project-lite",
- "rand 0.8.6",
- "slab",
- "tokio",
- "tokio-util",
- "tower-layer",
- "tower-service",
- "tracing",
-]
-
 [[package]]
 name = "tower"
 version = "0.5.3"
@@ -3197,13 +1444,13 @@ version = "0.6.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4cfcf7e2740e6fc6d4d688b4ef00650406bb94adf4731e43c096c3a19fe40840"
 dependencies = [
- "bitflags 2.11.1",
+ "bitflags",
  "bytes",
  "futures-util",
  "http",
  "http-body",
  "pin-project-lite",
- "tower 0.5.3",
+ "tower",
  "tower-layer",
  "tower-service",
  "url",
@@ -3240,7 +1487,7 @@ checksum = "7490cfa5ec963746568740651ac6781f701c9c5ea257c58e057f3ba8cf69e8da"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -3288,18 +1535,6 @@ version = "0.2.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e421abadd41a4225275504ea4d6566923418b7f05506fbc9c0fe86ba7396114b"
 
-[[package]]
-name = "twox-hash"
-version = "2.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9ea3136b675547379c4bd395ca6b938e5ad3c3d20fad76e7fe85f9e0d011419c"
-
-[[package]]
-name = "typenum"
-version = "1.20.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "40ce102ab67701b8526c123c1bab5cbe42d7040ccfd0f64af1a385808d2f43de"
-
 [[package]]
 name = "unicode-ident"
 version = "1.0.24"
@@ -3330,24 +1565,12 @@ dependencies = [
  "serde",
 ]
 
-[[package]]
-name = "utf8-ranges"
-version = "1.0.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7fcfc827f90e53a02eaef5e535ee14266c1d569214c6aa70133a624d8a3164ba"
-
 [[package]]
 name = "utf8_iter"
 version = "1.0.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6c140620e7ffbb22c2dee59cafe6084a59b5ffc27a8859a5f0d494b5d52b6be"
 
-[[package]]
-name = "utf8parse"
-version = "0.2.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"
-
 [[package]]
 name = "uuid"
 version = "1.23.1"
@@ -3356,7 +1579,6 @@ checksum = "ddd74a9687298c6858e9b88ec8935ec45d22e8fd5e6394fa1bd4e99a87789c76"
 dependencies = [
  "getrandom 0.4.2",
  "js-sys",
- "serde_core",
  "wasm-bindgen",
 ]
 
@@ -3366,18 +1588,6 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ba73ea9cf16a25df0c8caa16c51acb937d5712a8429db78a3ee29d5dcacd3a65"
 
-[[package]]
-name = "vcpkg"
-version = "0.2.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
-
-[[package]]
-name = "version_check"
-version = "0.9.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
-
 [[package]]
 name = "want"
 version = "0.3.1"
@@ -3453,7 +1663,7 @@ dependencies = [
  "bumpalo",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
  "wasm-bindgen-shared",
 ]
 
@@ -3483,33 +1693,20 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bb0e353e6a2fbdc176932bbaab493762eb1255a7900fe0fea1a2f96c296cc909"
 dependencies = [
  "anyhow",
- "indexmap 2.14.0",
+ "indexmap",
  "wasm-encoder",
  "wasmparser",
 ]
 
-[[package]]
-name = "wasm-streams"
-version = "0.4.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "15053d8d85c7eccdbefef60f06769760a563c7f0a9d6902a13d35c7800b0ad65"
-dependencies = [
- "futures-util",
- "js-sys",
- "wasm-bindgen",
- "wasm-bindgen-futures",
- "web-sys",
-]
-
 [[package]]
 name = "wasmparser"
 version = "0.244.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "47b807c72e1bac69382b3a6fb3dbe8ea4c0ed87ff5629b8685ae6b9a611028fe"
 dependencies = [
- "bitflags 2.11.1",
+ "bitflags",
  "hashbrown 0.15.5",
- "indexmap 2.14.0",
+ "indexmap",
  "semver",
 ]
 
@@ -3542,28 +1739,6 @@ dependencies = [
  "rustls-pki-types",
 ]
 
-[[package]]
-name = "winapi"
-version = "0.3.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419"
-dependencies = [
- "winapi-i686-pc-windows-gnu",
- "winapi-x86_64-pc-windows-gnu",
-]
-
-[[package]]
-name = "winapi-i686-pc-windows-gnu"
-version = "0.4.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
-
-[[package]]
-name = "winapi-x86_64-pc-windows-gnu"
-version = "0.4.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
-
 [[package]]
 name = "windows-core"
 version = "0.62.2"
@@ -3585,7 +1760,7 @@ checksum = "053e2e040ab57b9dc951b72c264860db7eb3b0200ba345b4e4c3b14f67855ddf"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -3596,7 +1771,7 @@ checksum = "3f316c4a2570ba26bbec722032c4099d8c8bc095efccdc15688708623367e358"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -3605,17 +1780,6 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f0805222e57f7521d6a62e36fa9163bc891acd422f971defe97d64e70d0a4fe5"
 
-[[package]]
-name = "windows-registry"
-version = "0.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "02752bf7fbdcce7f2a27a742f798510f3e5ad88dbe84871e5168e2120c3d5720"
-dependencies = [
- "windows-link",
- "windows-result",
- "windows-strings",
-]
-
 [[package]]
 name = "windows-result"
 version = "0.4.1"
@@ -3634,15 +1798,6 @@ dependencies = [
  "windows-link",
 ]
 
-[[package]]
-name = "windows-sys"
-version = "0.48.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "677d2418bec65e3338edb076e806bc1ec15693c5d0104683f2efe857f61056a9"
-dependencies = [
- "windows-targets 0.48.5",
-]
-
 [[package]]
 name = "windows-sys"
 version = "0.52.0"
@@ -3679,21 +1834,6 @@ dependencies = [
  "windows-link",
 ]
 
-[[package]]
-name = "windows-targets"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9a2fa6e2155d7247be68c096456083145c183cbbbc2764150dda45a87197940c"
-dependencies = [
- "windows_aarch64_gnullvm 0.48.5",
- "windows_aarch64_msvc 0.48.5",
- "windows_i686_gnu 0.48.5",
- "windows_i686_msvc 0.48.5",
- "windows_x86_64_gnu 0.48.5",
- "windows_x86_64_gnullvm 0.48.5",
- "windows_x86_64_msvc 0.48.5",
-]
-
 [[package]]
 name = "windows-targets"
 version = "0.52.6"
@@ -3727,12 +1867,6 @@ dependencies = [
  "windows_x86_64_msvc 0.53.1",
 ]
 
-[[package]]
-name = "windows_aarch64_gnullvm"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2b38e32f0abccf9987a4e3079dfb67dcd799fb61361e53e2882c3cbaf0d905d8"
-
 [[package]]
 name = "windows_aarch64_gnullvm"
 version = "0.52.6"
@@ -3745,12 +1879,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a9d8416fa8b42f5c947f8482c43e7d89e73a173cead56d044f6a56104a6d1b53"
 
-[[package]]
-name = "windows_aarch64_msvc"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dc35310971f3b2dbbf3f0690a219f40e2d9afcf64f9ab7cc1be722937c26b4bc"
-
 [[package]]
 name = "windows_aarch64_msvc"
 version = "0.52.6"
@@ -3763,12 +1891,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b9d782e804c2f632e395708e99a94275910eb9100b2114651e04744e9b125006"
 
-[[package]]
-name = "windows_i686_gnu"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a75915e7def60c94dcef72200b9a8e58e5091744960da64ec734a6c6e9b3743e"
-
 [[package]]
 name = "windows_i686_gnu"
 version = "0.52.6"
@@ -3793,12 +1915,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fa7359d10048f68ab8b09fa71c3daccfb0e9b559aed648a8f95469c27057180c"
 
-[[package]]
-name = "windows_i686_msvc"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f55c233f70c4b27f66c523580f78f1004e8b5a8b659e05a4eb49d4166cca406"
-
 [[package]]
 name = "windows_i686_msvc"
 version = "0.52.6"
@@ -3811,12 +1927,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1e7ac75179f18232fe9c285163565a57ef8d3c89254a30685b57d83a38d326c2"
 
-[[package]]
-name = "windows_x86_64_gnu"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "53d40abd2583d23e4718fddf1ebec84dbff8381c07cae67ff7768bbf19c6718e"
-
 [[package]]
 name = "windows_x86_64_gnu"
 version = "0.52.6"
@@ -3829,12 +1939,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9c3842cdd74a865a8066ab39c8a7a473c0778a3f29370b5fd6b4b9aa7df4a499"
 
-[[package]]
-name = "windows_x86_64_gnullvm"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0b7b52767868a23d5bab768e390dc5f5c55825b6d30b86c844ff2dc7414044cc"
-
 [[package]]
 name = "windows_x86_64_gnullvm"
 version = "0.52.6"
@@ -3847,12 +1951,6 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0ffa179e2d07eee8ad8f57493436566c7cc30ac536a3379fdf008f47f6bb7ae1"
 
-[[package]]
-name = "windows_x86_64_msvc"
-version = "0.48.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ed94fce61571a4006852b7389a063ab983c02eb1bb37b47f8272ce92d06d9538"
-
 [[package]]
 name = "windows_x86_64_msvc"
 version = "0.52.6"
@@ -3899,9 +1997,9 @@ checksum = "b7c566e0f4b284dd6561c786d9cb0142da491f46a9fbed79ea69cdad5db17f21"
 dependencies = [
  "anyhow",
  "heck",
- "indexmap 2.14.0",
+ "indexmap",
  "prettyplease",
- "syn",
+ "syn 2.0.117",
  "wasm-metadata",
  "wit-bindgen-core",
  "wit-component",
@@ -3917,7 +2015,7 @@ dependencies = [
  "prettyplease",
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
  "wit-bindgen-core",
  "wit-bindgen-rust",
 ]
@@ -3929,8 +2027,8 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9d66ea20e9553b30172b5e831994e35fbde2d165325bec84fc43dbf6f4eb9cb2"
 dependencies = [
  "anyhow",
- "bitflags 2.11.1",
- "indexmap 2.14.0",
+ "bitflags",
+ "indexmap",
  "log",
  "serde",
  "serde_derive",
@@ -3949,7 +2047,7 @@ checksum = "ecc8ac4bc1dc3381b7f59c34f00b67e18f910c2c0f50015669dde7def656a736"
 dependencies = [
  "anyhow",
  "id-arena",
- "indexmap 2.14.0",
+ "indexmap",
  "log",
  "semver",
  "serde",
@@ -3984,30 +2082,10 @@ checksum = "de844c262c8848816172cef550288e7dc6c7b7814b4ee56b3e1553f275f1858e"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
  "synstructure",
 ]
 
-[[package]]
-name = "zerocopy"
-version = "0.8.48"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "eed437bf9d6692032087e337407a86f04cd8d6a16a37199ed57949d415bd68e9"
-dependencies = [
- "zerocopy-derive",
-]
-
-[[package]]
-name = "zerocopy-derive"
-version = "0.8.48"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "70e3cd084b1788766f53af483dd21f93881ff30d7320490ec3ef7526d203bad4"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
-]
-
 [[package]]
 name = "zerofrom"
 version = "0.1.8"
@@ -4025,7 +2103,7 @@ checksum = "11532158c46691caf0f2593ea8358fed6bbf68a0315e80aae9bd41fbade684a1"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
  "synstructure",
 ]
 
@@ -4065,7 +2143,7 @@ checksum = "625dc425cab0dca6dc3c3319506e6593dcb08a9f387ea3b284dbd52a92c40555"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn",
+ "syn 2.0.117",
 ]
 
 [[package]]
@@ -4073,31 +2151,3 @@ name = "zmij"
 version = "1.0.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"
-
-[[package]]
-name = "zstd"
-version = "0.13.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e91ee311a569c327171651566e07972200e76fcfe2242a4fa446149a3881c08a"
-dependencies = [
- "zstd-safe",
-]
-
-[[package]]
-name = "zstd-safe"
-version = "7.2.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f49c4d5f0abb602a93fb8736af2a4f4dd9512e36f7f570d66e65ff867ed3b9d"
-dependencies = [
- "zstd-sys",
-]
-
-[[package]]
-name = "zstd-sys"
-version = "2.0.16+zstd.1.5.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "91e19ebc2adc8f83e43039e79776e3fda8ca919132d68a1fed6a5faca2683748"
-dependencies = [
- "cc",
- "pkg-config",
-]
diff --git a/Cargo.toml b/Cargo.toml
index 59e6f65..5472d3a 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,30 +1,35 @@
 [package]
 name = "reasonkit-think-mcp"
-version = "0.1.3"
+version = "0.2.0"
 edition = "2024"
 # Match repo `rust-toolchain.toml` (latest stable); `rmcp` is edition-2024-friendly.
 rust-version = "1.95"
 license = "Apache-2.0"
 authors = ["Len P. van der Hof <Lenvanderhof@ReasonKit.sh>"]
-description = "Rust MCP server for auditable sequential and tree-of-thoughts reasoning, aligned with ReasonKit skills."
+description = "Agent-native MCP server for evidence-gated reasoning, fail-closed verification, and exportable decision audits."
 repository = "https://github.com/reasonkit/reasonkit-think"
 documentation = "https://docs.rs/reasonkit-think-mcp"
 homepage = "https://reasonkit.sh"
 readme = "README.md"
 keywords = ["mcp", "reasoning", "llm", "agents", "tree-of-thoughts"]
 categories = ["development-tools"]
-exclude = ["target/", "scripts/__pycache__/"]
+include = [
+    "/src/**",
+    "/Cargo.toml",
+    "/Cargo.lock",
+    "/README.md",
+    "/CHANGELOG.md",
+    "/LICENSE",
+]
 
 [dependencies]
 # Official MCP Rust SDK: https://github.com/modelcontextprotocol/rust-sdk — see docs/MCP_STACK_RESEARCH.md
-rmcp = { version = "1.6.0", default-features = false, features = [
+rmcp = { version = "3.1.4", default-features = false, features = [
     "server",
     "macros",
     "schemars",
     "transport-io",
 ] }
-# Published crates.io dep (path `../reasonkit-mem` for umbrella dev checkouts).
-reasonkit-mem = { version = "0.1.7", default-features = false, features = ["compression"] }
 tokio = { version = "1.52", features = ["io-std", "macros", "rt", "rt-multi-thread"] }
 anyhow = "1.0"
 serde = { version = "1.0", features = ["derive"] }
diff --git a/README.md b/README.md
index a2a1ed0..8f9cd43 100644
--- a/README.md
+++ b/README.md
@@ -1,285 +1,241 @@
 <p align="center">
-  <img src=".brand/assets/ReasonKit-think_logo_banner.png" alt="ReasonKit-think — Advanced Reasoning MCP for AI Coding Agents" width="100%"/>
+  <img src=".brand/assets/ReasonKit-think_logo_banner.png" alt="ReasonKit Think — auditable decision protocol for MCP agents" width="100%"/>
 </p>
 
-# 🧠 ReasonKit Think MCP
+# ReasonKit Think MCP
 
 <p align="center">
-  <strong>Advanced Reasoning MCP for AI Coding Agents</strong><br/>
-  <i>CoT + ToT + GoT + Verification + Governance Routing</i>
+  <strong>Auditable decisions and fail-closed evidence gates for MCP agents</strong><br/>
+  <i>Designed, Not Dreamed. Turn prompts into protocols.</i>
 </p>
 
 <p align="center">
-  <img alt="Rust" src="https://badges.reasonkit.sh/static/v1?label=Rust&message=1.95%2B&color=ember&logo=rust"/>
-  <img alt="MCP" src="https://badges.reasonkit.sh/static/v1?label=MCP&message=rmcp&color=teal"/>
-  <img alt="License" src="https://badges.reasonkit.sh/static/v1?label=license&message=Apache%202.0&color=pro"/>
-  <img alt="Status" src="https://badges.reasonkit.sh/static/v1?label=status&message=v1%20%2B%20v2%20live&color=ok"/>
-  <a href="https://github.com/reasonkit/reasonkit-think/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/reasonkit/reasonkit-think/actions/workflows/ci.yml/badge.svg"/></a>
+  <a href="https://github.com/reasonkit/ReasonKit-think/actions/workflows/ci.yml">
+    <img src="https://badges.reasonkit.sh/github/actions/workflow/status/reasonkit/ReasonKit-think/ci.yml?branch=main&logo=github&label=CI&color=accent" alt="CI status" />
+  </a>
+  <a href="https://crates.io/crates/reasonkit-think-mcp">
+    <img src="https://badges.reasonkit.sh/crates/v/reasonkit-think-mcp?logo=rust&color=ok" alt="Crates.io version" />
+  </a>
+  <a href="https://docs.rs/reasonkit-think-mcp">
+    <img src="https://badges.reasonkit.sh/docsrs/reasonkit-think-mcp?logo=docs.rs&color=accent" alt="docs.rs" />
+  </a>
+  <a href="https://crates.io/crates/reasonkit-think-mcp">
+    <img src="https://badges.reasonkit.sh/crates/d/reasonkit-think-mcp?color=signal&logo=rust" alt="Downloads" />
+  </a>
+  <img alt="Rust 1.95+" src="https://badges.reasonkit.sh/static/v1?label=Rust&message=1.95%2B&color=ember&logo=rust" />
+  <img alt="MCP via rmcp" src="https://badges.reasonkit.sh/static/v1?label=MCP&message=rmcp&color=teal" />
+  <a href="LICENSE">
+    <img src="https://badges.reasonkit.sh/static/v1?label=license&message=Apache%202.0&color=pro" alt="Apache-2.0" />
+  </a>
+  <a href="https://github.com/reasonkit/ReasonKit-think">
+    <img src="https://badges.reasonkit.sh/static/v1?label=status&message=v0.2%20release%20candidate&color=lease" alt="v0.2 release candidate" />
+  </a>
 </p>
 
-<p align="center">
-  <a href="https://github.com/reasonkit/reasonkit-think"><strong>GitHub</strong></a>
-  ·
-  <a href="https://crates.io/crates/reasonkit-think-mcp"><strong>Crates.io</strong></a>
-  ·
-  <a href="https://reasonkit.sh"><strong>reasonkit.sh</strong></a>
-</p>
+ReasonKit Think is a Rust stdio MCP server that helps a host agent turn a
+decision into an inspectable graph, test critical claims against supplied
+evidence, stop when required evidence is missing, and export an audit artifact.
+
+The host agent remains the semantic engine. ReasonKit Think does not expose a
+model's private chain of thought and does not invent model-backed or
+retrieval-backed findings. It records only content explicitly submitted through
+its tools, labels deterministic heuristics, and fails closed at unsupported
+provider boundaries.
 
----
+- GitHub: <https://github.com/reasonkit/ReasonKit-think>
+- Crate: <https://crates.io/crates/reasonkit-think-mcp>
+- MCP Registry name: `mcp-name: io.github.reasonkit/reasonkit-think`
 
-## 🎯 Why ReasonKit Think?
+## Quick start
 
-Most AI agents today either reason shallowly and move too fast (causing costly mistakes) or reason deeply but become hard to debug, blindly trusting their own hallucinations.
+### 1. Install the binary on PATH
 
-**ReasonKit Think** bridges this gap. It provides a structured, auditable, and mathematically grounded reasoning engine directly to your favorite MCP-compatible AI clients (Cursor, Claude Code, Gemini CLI, Copilot).
+Prerequisites: Rust 1.95+ and an MCP-compatible host.
 
-We transform abstract AI "thinking" into concrete, inspectable, and governable **Protocols**.
+```sh
+git clone https://github.com/reasonkit/ReasonKit-think.git
+cd ReasonKit-think
+cargo install --locked --path .
+reasonkit-think-mcp --version
+reasonkit-think-mcp doctor --json
+```
 
-### The ReasonKit Advantage:
-* **Anti-Hallucination:** Strict evidence-based verification gates. Claims without proof are marked `DATA_DEFICIT` and fail closed.
-* **Traceable Auditability:** Every thought, branch, and decision is logged in a stateful DAG (Directed Acyclic Graph) for post-mortem analysis.
-* **Dynamic Paradigms:** Seamlessly switch between Chain-of-Thought (CoT), Tree-of-Thoughts (ToT), and Graph-of-Thoughts (GoT) on the fly.
-* **Skeleton-of-Thought Phases:** For multi-aspect problems, route into outline-first skeleton branches, elaborate them independently, then merge through GoT checkpoints.
-* **Honest Provider Boundaries:** Heuristic outputs are labeled as heuristic; model-backed and retrieval-backed analysis fail closed until real provider bindings are configured and implemented.
-* **State Survival & Observability:** Local snapshots survive process restarts, and telemetry resources expose tool latency, verification outcomes, pruning, and graph-depth metrics.
-* **Assumption-Led Reasoning:** Critical premises are tracked in an assumption ledger and can block convergence until resolved.
-* **Replayable Convergence:** Agents can replay a session timeline, run checkpoints, and converge only after quality, evidence, and assumption gates pass.
+This source-checkout command is the presently verifiable `0.2.0` release
+candidate path. The Registry metadata is not live yet; after v0.2.0 publication
+the equivalent crates.io install will be:
 
----
+```sh
+cargo install --locked --version 0.2.0 reasonkit-think-mcp
+```
 
-## 🧠 Thinking Modes
+Do not use an unversioned crates.io install to evaluate this release candidate:
+it currently resolves an older published release. `cargo install` places the
+executable in Cargo's binary directory, normally `~/.cargo/bin`; add that
+directory to `PATH` if the version check cannot find it.
 
-ReasonKit Think now uses accessible user-facing Thinking Modes. The academic method names are still supported internally for compatibility, but most users should start with these names:
+### 2. Register the stdio server
 
-| Mode | Internal behavior | Cost | Best for | Aliases |
-| --- | --- | --- | --- | --- |
-| **Auto** | Hybrid router / adaptive mode selector | Medium | Default entry point for most work | Autopilot, Guided, Smart, Adaptive, Flow, Default |
-| **Quick** | CoT-style concise reasoning audit | Low | Fast first pass, simple debugging, concise review | Sprint, Simple, Linear, First Pass, Baseline, Trace |
-| **Explore** | Tree-of-Thoughts branching search | High | Options, hypotheses, tradeoffs, comparisons | Branch, Diverse, Brainstorm, Compare, Scenario, Fork |
-| **Map** | Graph-of-Thoughts synthesis | High | Dependencies, conflicts, refinements, merged plans | Architect, Synthesis, Weave, Network, Merge, System |
-| **Sketch** | Skeleton-of-Thought outline-first expansion | Medium | Multi-section answers, plans, or independent lenses | Outline, Blueprint, Scaffold, Skeleton, Draft, Frame |
-| **Test** | CoVe + ReasonKit validation pipeline | Medium to high | Verification, release gates, go/no-go decisions | Verify, Audit, Proof, Guard, Red Team, Check |
+Use the executable name, not a machine-specific absolute path.
 
-**Auto is the default.** Tell ReasonKit what you need; it chooses the right reasoning path and returns an auditable result.
+```sh
+codex mcp add reasonkit-think -- reasonkit-think-mcp
+claude mcp add --scope user reasonkit-think -- reasonkit-think-mcp
+gemini mcp add --scope user reasonkit-think reasonkit-think-mcp
+copilot mcp add reasonkit-think -- reasonkit-think-mcp
+```
 
-Recommended first tool:
+For Cursor and other JSON-configured hosts:
 
 ```json
 {
-  "tool": "run_thinking_mode",
-  "arguments": {
-    "mode": "Auto",
-    "intent": "Compare the migration options, verify risky claims, and return an auditable recommendation."
+  "mcpServers": {
+    "reasonkit-think": {
+      "type": "stdio",
+      "command": "reasonkit-think-mcp",
+      "args": []
+    }
   }
 }
 ```
 
-`Test` means claim verification and governance; it does not run your project test suite.
-
-Clients can discover the full machine-readable mode contract through the native
-MCP resource `reasoning://thinking-modes`, including aliases, internal method
-mapping, estimated cost, required behavior, safety rules, and recommended tools.
-
-> **Example Prompt:**
-> *"Use Auto mode with reasonkit-think. Compare the migration options, verify risky claims, and return an auditable recommendation with evidence gaps."*
+See [client-specific setup](docs/clients/) and the reusable
+[configuration examples](examples/client-configs/).
 
-The compatibility mapping is straightforward:
+### 3. Run the auditable decision path
 
-- `Quick` maps to `cot`-style linear audit output.
-- `Explore` maps to `tot` branching, scoring, and pruning.
-- `Map` maps to `got` graph dependencies and synthesis.
-- `Sketch` uses a `got` session plus Skeleton-of-Thought tools.
-- `Test` uses the ReasonKit verification and governance pipeline.
+Ask your host agent:
 
----
+> Use reasonkit-think Auto mode. Compare the options, state the critical
+> assumptions, author the decision nodes, attach explicit evidence to risky
+> claims, stop if a critical gap remains, and return the decision plus its
+> audit artifact.
 
-## 🛡️ The Verification & Governance Engine
+The complete core path is:
 
-Reasoning is useless if it's based on fabricated facts. ReasonKit Think enforces **CoVe (Chain of Verification)** and **Triangulation**.
+1. `run_thinking_mode` — start in Auto and create the working decision graph.
+2. `add_thought_node` — replace or refine a returned scaffold with the host
+   agent's actual analysis. Consensus never promotes a heuristic scaffold or
+   the original goal into a final answer.
+3. `verify_thoughts` — submit explicit claims and evidence; missing evidence is
+   `DATA_DEFICIT`, not a guessed success.
+4. `run_reasoning_checkpoint` — evaluate blockers and readiness.
+5. `consensus_answer` — produce a decision only when the configured gates allow
+   it and an agent-authored candidate is on the frontier.
+6. `export_reasoning_audit` — return the graph, verification matrix, route, and
+   recorded limitations.
 
-When you trigger a deep deliberation, the server builds a **Verification Matrix** from claims you supply (or that the **host agent** derives and passes to `verify_thoughts`). The MCP server does not run an LLM to extract claims from thought text.
-1. **Registers Claims:** Records factual or architectural assertions with explicit evidence tiers (via `verify_thoughts` / thinking-mode `Test`).
-2. **Evaluates Evidence:** Applies tier rules and triangulation heuristics to each claim.
-3. **Fails Closed:** If a critical claim has conflicting sources (`SOURCE_CONFLICT`) or missing data (`DATA_DEFICIT`), governance routes return `GATHER_MORE_EVIDENCE` instead of unsafe conclusions.
+Use `record_assumption` before verification when a premise must remain visible
+and block the decision until resolved.
 
-**Agent-native protocol:** `expand_thoughts` emits labeled **scaffolds**; the host agent must replace them with substantive nodes via `add_thought_node`. Responses include `next_action` coaching, optional ReAct nodes (`record_reasoning_action` / `record_reasoning_observation`), structural `mcts_select_path`, and cross-session `query_failure_memory`.
+## Tool packs
 
----
+`REASONKIT_TOOL_PACK` controls discovery breadth without changing stored data.
 
-## 🚀 Quickstart & Installation
-
-**Prerequisites:** Rust 1.95+ and an `rmcp`-compatible MCP host.
-
-### Install via Crates.io (Recommended)
-```bash
-cargo install reasonkit-think-mcp
-```
-
-### Build from Source
-```bash
-git clone https://github.com/reasonkit/reasonkit-think.git
-cd reasonkit-think
-export CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-$PWD/target}"
-cargo build --release
-```
-
-To run a quick health check of the server and its tool surface:
-```bash
-./scripts/smoke_test.py
-```
-
-To verify the release binary, protocol handshake, schema compatibility, and
-local Copilot/Cursor/Claude config paths:
-```bash
-./scripts/client_compat_check.py
-```
-
-### Docker
-
-The image packages the stdio MCP server for supervised MCP hosts:
-```bash
-docker build -t reasonkit-think-mcp .
-docker run --rm -i reasonkit-think-mcp
-```
+| Pack | Tools | Intended use |
+| --- | ---: | --- |
+| `core` | 13 | Default. Complete route → deliberate → evidence → gate → decision → audit path. |
+| `standard` | 33 | Adds graph, scoring, pruning, replay, quality, and Skeleton-of-Thought operations. |
+| `full` | 47 | compatibility escape hatch for existing clients that depend on the complete legacy inventory. |
 
----
+Keep the default `core` pack for new installations. Set the environment variable
+on the MCP server only when a client needs a broader surface:
 
-## ⚙️ MCP Client Configuration
-
-Add `reasonkit-think-mcp` to your favorite AI IDE or CLI. **Ensure you provide the ABSOLUTE PATH to the release binary.** If you built from source with the default command above, that path is usually:
-
-```text
-/path/to/reasonkit-think/target/release/reasonkit-think-mcp
-```
-
-<details>
-<summary><strong>Cursor & Cursor Agent</strong></summary>
-
-Add under `mcpServers` in `~/.cursor/mcp.json` or `.cursor/mcp.json`:
 ```json
 {
   "mcpServers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
+      "command": "reasonkit-think-mcp",
       "args": [],
-      "env": { "TMPDIR": "/tmp" }
+      "env": { "REASONKIT_TOOL_PACK": "full" }
     }
   }
 }
 ```
-Reload MCP from Cursor settings.
-</details>
-
-<details>
-<summary><strong>Claude Code</strong></summary>
 
-Add to your `~/.claude.json`:
-```json
-"reasonkit-think": {
-  "type": "stdio",
-  "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
-  "args": [],
-  "env": { "TMPDIR": "/tmp" }
-}
+The exact lexical inventories are versioned in
+[`evals/tool-packs.json`](evals/tool-packs.json) and enforced in CI.
+
+## Thinking modes
+
+Auto is the recommended entry point. The other names make routing intent
+explicit while retaining compatibility with the underlying reasoning methods.
+
+| Mode | Behavior | Best for |
+| --- | --- | --- |
+| **Auto** | Selects an appropriate supported path | Most decisions |
+| **Quick** | Concise linear audit | First passes and simple reviews |
+| **Explore** | Branch, score, and prune | Options and competing hypotheses |
+| **Map** | Link dependencies and synthesize | Architecture and system decisions |
+| **Sketch** | Outline, elaborate, then merge | Multi-part plans and documents |
+| **Test** | Verification and governance pipeline | Release gates and go/no-go decisions |
+
+`Test` validates supplied claims; it does not run a repository's test suite.
+The machine-readable mode catalog is available at
+`reasoning://thinking-modes`.
+
+## Evidence and audit contract
+
+- Claims and evidence come from the host agent or user. The server does not run
+  an LLM to extract facts from hidden reasoning.
+- Evidence tiers, independence groups, contradictions, and ambiguity are
+  evaluated deterministically.
+- Critical missing or conflicting evidence routes to `GATHER_MORE_EVIDENCE`.
+- Unresolved high/critical assumptions block convergence under the default
+  policy.
+- Heuristic scores are structural proxies with provenance and limitations; they
+  are not semantic confidence.
+- Mutable tool calls persist submitted state locally so sessions can be replayed
+  and audited.
+- The active transport is stdio. No daemon, hosted service, or telemetry export
+  is required.
+
+ReasonKit Think cannot read a repository, URL, or `@mention` by itself. Have the
+host use its normal filesystem/browser tools, then pass the relevant evidence
+packet into ReasonKit Think.
+
+## Development and verification
+
+```sh
+just check       # formatting, clippy, Rust tests, contract unit tests
+just build       # release binary
+just eval        # exact pack/protocol matrix plus planted fail-closed cases
+just audit       # deny vulnerabilities and advisory warnings
+just package-check  # build and verify the publishable crate archive
+just registry-check # validate server.json against the pinned Registry schema
+just dist-check     # validate the workflow-free cargo-dist manifest
 ```
-Restart your Claude Code session.
-</details>
-
-<details>
-<summary><strong>Other Clients (Gemini CLI, Codex, Copilot, OpenCode)</strong></summary>
-
-For other clients, configure a standard `stdio` MCP server pointing to the absolute path of `reasonkit-think-mcp`. See the `docs/clients/` folder for comprehensive setup guides for all major AI coding platforms.
-</details>
-
-Reusable JSON snippets are available in `examples/client-configs/`.
-
-Workflow request examples are available in `examples/`:
 
-* `run-thinking-mode-auto.json`
-* `run-thinking-mode-explore.json`
-* `run-thinking-mode-sketch.json`
-* `run-thinking-mode-test.json`
-* `feature-triage-auto-map.json`
+The full-pack end-to-end smoke path is:
 
-If a client reports `ENOENT` or "failed to connect", rebuild the release binary and rerun the compatibility check:
-
-```bash
-export CARGO_TARGET_DIR="$PWD/target"
-cargo build --release
-./scripts/client_compat_check.py
+```sh
+python3 scripts/smoke_test.py
 ```
 
----
-
-## 🗣️ Natural Wording Triggers
-
-You do not have to remember exact MCP tool names. `run_thinking_mode`,
-`reasoning_intent_router`, and `reasoning_autopilot` understand both the new
-mode names and legacy method names:
-
-* **Auto:** `"auto"`, `"autopilot"`, `"guided"`, `"choose for me"`, `"default"`
-* **Quick:** `"quick"`, `"simple"`, `"sprint"`, `"linear reasoning"`, `"first pass"`
-* **Explore:** `"explore options"`, `"branch"`, `"brainstorm"`, `"compare"`, `"fork"`
-* **Map:** `"map"`, `"architect"`, `"synthesis"`, `"merge paths"`, `"network"`
-* **Sketch:** `"sketch"`, `"outline first"`, `"blueprint"`, `"scaffold"`, `"skeleton"`
-* **Test:** `"test"`, `"verify"`, `"audit"`, `"proof"`, `"red team"`, `"check"`
-* **Support layer:** `"assumption"`, `"checkpoint"`, `"quality gate"`, `"replay"`, `"memory snapshot"`, `"go/no-go"`
-
-Use `docs/README.md` for the full vocabulary, mode workflows, and tool map.
-
-Common client argument shapes are accepted: arrays may be passed directly or as
-JSON-stringified arrays, numeric fields may be numeric strings, boolean fields
-may be boolean-like strings, and graph tools accept friendly node references
-such as `frontier`, `root`, and 1-based indexes like `"1"`.
+Architecture, tools, prompts, and resources are indexed in
+[`docs/README.md`](docs/README.md). The deterministic runtime contract is
+described in [`evals/README.md`](evals/README.md).
 
-ReasonKit Think does not inspect local files, repositories, URLs, or `@mentions`
-by itself. For requests such as “rank features from `features/todo/`”, first use
-your client’s filesystem/search tools to read the files, then pass the extracted
-feature packet into `run_thinking_mode` Auto and Map mode. The workflow resource
-`reasoning://workflows/feature-triage` is built for that pattern.
+## Distribution status
 
----
+`server.json` and `dist-workspace.toml` describe the `0.2.0` release candidate.
+This is not yet a live Registry record: crates.io publication must complete
+first, the Registry manifest must resolve that exact package version, and the
+post-publish validation must pass. The cargo-dist file is configuration only;
+this repository intentionally has no active release workflow, tag, or publish
+automation in this tranche.
 
-## Production Contract
+## Community
 
-ReasonKit Think is explicit about what it does and does not do:
+See [CONTRIBUTING.md](CONTRIBUTING.md) for development and contract checks,
+[SECURITY.md](SECURITY.md) for private vulnerability reporting, and
+[features/README.md](features/README.md) for the machine-checked planning-record
+inventory.
 
-* `heuristic` outputs are deterministic structural proxies with provenance and limitations.
-* `model_backed` and `retrieval` provider modes are exposed as fail-closed boundaries in `reasoning://config/providers`; the server does not manufacture semantic confidence when no real provider binding exists.
-* `record_assumption` and `set_assumption_status` expose a first-class assumption ledger; unresolved high/critical assumptions gate consensus and convergence.
-* `run_reasoning_checkpoint`, `replay_reasoning_session`, and `converge_reasoning` make readiness, replayability, and final synthesis explicit MCP operations.
-* `apply_reasoning_pattern` gives agents reusable reasoning modes such as debugging, scientific method, systems mapping, dialectical synthesis, and code reasoning.
-* `start_skeleton_of_thought` and `run_skeleton_elaboration` implement a routed SoT workflow. Unsuitable tasks are blocked by default, heuristic mode creates scaffolds/prompt packs only, and model-backed mode fails closed until a real provider exists.
-* `plan_tool_sequence` returns stage-aware tool call templates; it plans but never executes tools on behalf of the caller.
-* `export_memory_snapshot` returns a memory-ready payload. External sinks are explicit `boundary_unavailable` responses unless a caller persists the payload separately.
-* `run_thinking_mode` is the primary operational mode runner for Auto, Quick, Explore, Map, Sketch, and Test.
-* `reasoning://thinking-modes` exposes the public Thinking Mode contract used by router/autopilot/run outputs.
-* Mutable tools persist state to `~/.local/share/sh.reasonkit.think/state.json` after releasing store locks.
-* `reasoning://telemetry/summary` reports in-process counters for MCP clients, alongside OpenTelemetry metric instruments.
-* The active transport is `stdio`; `reasoning://config/transports` reports streamable HTTP as not active in this build.
+## License
 
----
-
-## 📜 Showcase Prompt (Try this!)
-
-Want to test the full power of ReasonKit Think? Copy and paste this to your agent:
-
-> *"Use the reasonkit-think reasoning system to think this through deeply and transparently. Our production database is experiencing 100% CPU utilization, but traffic hasn't spiked. We have 3 theories: a bad index from yesterday's migration, a runaway background job, or a hardware degradation.
->
-> Please approach this in a way that is highly auditable:
-> 1. Start with a quick first pass to clarify assumptions (CoT).
-> 2. Switch to deeper Tree-of-Thought (ToT) reasoning to generate specific diagnostic strategies for all 3 theories.
-> 3. Verify critical claims (e.g., 'we can safely kill the background job') with evidence quality in mind.
-> 4. Use Graph-of-Thoughts (GoT) to distill the branches into a single, concrete, prioritized action plan.
-> 5. Export a traceable reasoning audit showing why this plan is trustworthy."*
-
----
+Apache-2.0. See [LICENSE](LICENSE).
 
 <p align="center">
   <img src=".brand/assets/Powered_by_ReasonKit_logo_banner.png" alt="Powered by ReasonKit" width="100%"/>
 </p>
-
-**License:** Apache-2.0
-*Designed, Not Dreamed. Turn Prompts into Protocols.*
-[reasonkit.sh](https://reasonkit.sh)
diff --git a/SECURITY.md b/SECURITY.md
new file mode 100644
index 0000000..dc3a4fa
--- /dev/null
+++ b/SECURITY.md
@@ -0,0 +1,28 @@
+# Security Policy
+
+## Supported versions
+
+Security fixes target the default branch and, when practical, the latest
+published release. The `0.2.0` release candidate is not yet published; older
+releases receive fixes on a best-effort basis.
+
+## Report a vulnerability
+
+Please report vulnerabilities privately through
+[GitHub's private vulnerability form](https://github.com/reasonkit/ReasonKit-think/security/advisories/new).
+If that form is unavailable, email `Lenvanderhof@ReasonKit.sh` with the subject
+`ReasonKit Think security report`.
+
+Include the affected version or commit, impact, reproduction steps, and a
+minimal proof of concept. Redact credentials, personal data, and unrelated
+machine details. Do not disclose the report in a GitHub issue or discussion
+before a fix is coordinated.
+
+The maintainers will acknowledge the report, attempt to reproduce it, assess
+affected versions, and coordinate remediation and disclosure. Response times
+depend on severity and maintainer availability; no fixed SLA is promised.
+
+Relevant reports include MCP framing or schema bypasses, fail-open governance,
+unsafe local-state handling, command execution, credential exposure, and
+dependency vulnerabilities. Ordinary disagreement with a heuristic score is a
+quality issue unless it also crosses a documented security or policy boundary.
diff --git a/dist-workspace.toml b/dist-workspace.toml
new file mode 100644
index 0000000..455f9e4
--- /dev/null
+++ b/dist-workspace.toml
@@ -0,0 +1,17 @@
+[workspace]
+members = ["cargo:."]
+
+[dist]
+cargo-dist-version = "0.32.0"
+ci = []
+installers = ["shell", "powershell"]
+hosting = ["github"]
+targets = [
+  "aarch64-apple-darwin",
+  "aarch64-unknown-linux-gnu",
+  "x86_64-apple-darwin",
+  "x86_64-unknown-linux-gnu",
+  "x86_64-pc-windows-msvc",
+]
+install-path = "CARGO_HOME"
+install-updater = false
diff --git a/docs/ARCHITECTURE.md b/docs/ARCHITECTURE.md
index 49b1bcd..a3541ef 100644
--- a/docs/ARCHITECTURE.md
+++ b/docs/ARCHITECTURE.md
@@ -211,7 +211,7 @@ sequenceDiagram
   Client->>MCP: run_reasoning_checkpoint / converge_reasoning
   MCP->>Deliberation: gate convergence on assumptions, verification, quality
   Client->>MCP: run_reasonkit_pipeline
-  MCP->>Skills: emit selected heuristic stage markers
+  MCP->>Skills: emit structured stage contracts, gates, blockers, and agent tasks
   Client->>MCP: consensus_answer
   MCP-->>Client: answer + heuristic confidence + route + analysis metadata
 ```
@@ -256,7 +256,11 @@ Mandatory checks before `consensus_answer` success:
 ## Operational Resources
 
 - `reasoning://thinking-modes` reports the public Thinking Mode catalog and
-  compatibility mapping for clients.
+  compatibility mapping filtered to the active tool pack.
+- `reasoning://config/tool-packs` reports the startup-fixed `core`, `standard`,
+  or `full` router surface, exact membership, and the full-pack compatibility
+  rollback. The default core router exposes 13 of 47 registered routes; rmcp's
+  stored `ToolRouter` rejects disabled routes as well as hiding them from list.
 - `reasoning://config/providers` reports active heuristic mode, provider
   boundaries, required environment variables, and unavailable modes without
   leaking configured endpoint URLs.
@@ -277,6 +281,19 @@ Mandatory checks before `consensus_answer` success:
 - `reasoning://schemas/{name}` reports Vertex/Gemini-compatible JSON schemas for
   tools and resources.
 
+`doctor --json` treats an unset or `agent` `BACKEND_MODE` as healthy. Invalid
+values and the reserved `external` mode produce a secret-free warning and use
+the safe `agent` fallback because no external provider is bound.
+
+Successful results and handler-generated application errors returned through
+ReasonKit's common result helpers use MCP dual-format JSON: the same value is
+returned as `structuredContent` and as a pretty JSON text block for legacy
+clients. Protocol errors emitted by rmcp before a handler runs, including
+argument-deserialization failures, remain framework-generated text errors.
+The full-pack `think_query` name remains discoverable for compatibility, but
+returns `experimental_memory_unavailable` until a production ingestion and
+retrieval provider is explicitly bound; it never simulates a search.
+
 ## Security Controls
 
 - Existing prompt-injection scan/redaction remains enabled.
@@ -288,4 +305,5 @@ Mandatory checks before `consensus_answer` success:
 
 - Keep v1 tools stable and available.
 - New tools are additive; no breaking rename.
-- Documented protocol contract in `docs/research/SOURCES.md` and per-tool docs.
+- Documented protocol and ecosystem decisions in `docs/MCP_STACK_RESEARCH.md`
+  and per-tool docs.
diff --git a/docs/MCP_STACK_RESEARCH.md b/docs/MCP_STACK_RESEARCH.md
new file mode 100644
index 0000000..5032339
--- /dev/null
+++ b/docs/MCP_STACK_RESEARCH.md
@@ -0,0 +1,115 @@
+# MCP stack and product opportunity research
+
+- Status: v0.2 implementation basis
+- Reviewed: 2026-08-23
+- Scope: official MCP, Rust SDK, Registry, distribution, and adjacent reference
+  implementations
+
+## Product conclusion
+
+ReasonKit Think should not compete for the longest scratchpad. Its durable OSS
+position is the reasoning governance layer for agents: the host authors the
+semantic work, while the server records structure, applies evidence and
+authorship gates, routes unsafe conclusions away from `PROCEED`, and exports an
+artifact another human or agent can inspect.
+
+The v0.2 release candidate therefore prioritizes a small complete default
+surface, protocol compatibility, truthful provider boundaries, transactional
+governance, adversarial contract tests, and verifiable packaging. More reasoning
+paradigms are lower value until the daily decision workflow is easy to adopt.
+
+## Primary-source findings
+
+| Signal | Product implication | v0.2 response |
+| --- | --- | --- |
+| The official Sequential Thinking reference exposes one tool and describes a dynamic reflective problem-solving process. | One-tool discoverability is the adoption benchmark; copying the scratchpad does not create a wedge. | Default to a complete 13-tool core and keep 33/47-tool packs explicit. |
+| MCP tools can return `structuredContent`, while text content remains useful for compatibility. | Governance results should be machine-readable without abandoning older hosts. | Common handler results return the same JSON in structured and text forms; pre-handler SDK errors remain framework-owned. |
+| The official Rust SDK is the protocol implementation boundary and evolves with MCP revisions. | SDK drift is product and security debt, not merely dependency hygiene. | Upgrade to `rmcp` 3.1.4 and test five protocol revisions over stdio. |
+| The MCP Registry accepts Cargo package metadata and validates `server.json` against a published schema. | Discovery claims should be reproducible before any live Registry claim is made. | Validate the exact 0.2.0 package and Registry manifest in CI; do not claim publication yet. |
+| Cargo packages from the manifest include-list, and `cargo publish --dry-run` verifies the packaged crate. | The release artifact, not the worktree, is the supply-chain unit users receive. | Root-anchor the include-list and gate the exact ten-file archive. |
+| cargo-dist can generate native release artifacts from checked configuration. | Prebuilt binaries can remove Rust-toolchain friction, but release automation must not be speculative. | Validate a workflow-free cargo-dist manifest now; add publishing only with explicit release authority. |
+| MCP Apps provides a standard path for interactive UI attached to MCP tools. | A graph viewer can become a progressive enhancement without turning the project into a separate web product. | Defer to a read-only audit/graph viewer after the protocol is stable. |
+
+Primary references:
+
+- [MCP tools specification](https://modelcontextprotocol.io/specification/draft/server/tools)
+- [Official MCP Rust SDK](https://github.com/modelcontextprotocol/rust-sdk)
+- [Official Sequential Thinking server](https://github.com/modelcontextprotocol/servers/blob/main/src/sequentialthinking/README.md)
+- [MCP Registry quickstart](https://modelcontextprotocol.io/registry/quickstart)
+- [Registry Cargo package type](https://github.com/modelcontextprotocol/registry/blob/main/docs/modelcontextprotocol-io/package-types.mdx)
+- [Cargo package and publish reference](https://doc.rust-lang.org/cargo/reference/publishing.html)
+- [cargo-dist documentation](https://axodotdev.github.io/cargo-dist/)
+- [MCP Apps SDK](https://github.com/modelcontextprotocol/ext-apps)
+- [cargo-audit documentation](https://github.com/rustsec/rustsec/blob/main/cargo-audit/README.md)
+
+## Verified repository reality
+
+- The server has 47 compatibility tools, 11 prompts, and native resources.
+- The ten-stage ReasonKit pipeline already emits structured contracts, stage
+  gates, blockers, and host `agent_tasks`; it is not merely a list of labels.
+- Semantic model and retrieval execution remain deliberately unavailable unless
+  a real provider is bound. Heuristic output carries provenance.
+- The dominant engineering risk was concentrated state logic in `src/main.rs`
+  plus wire-level adoption friction, not a lack of reasoning modes.
+- The v0.2 contract tests cover exact tool-pack discovery, schema shape, five
+  protocol versions, and eight planted governance bypasses.
+
+## Sequenced opportunity map
+
+### v0.3: evidence-native coding workflows
+
+1. Add typed evidence packets for `git_diff`, `test_log`, `ci_check`,
+   `file_excerpt`, `issue_comment`, and `web_source`, with explicit provenance
+   and conservative defaults.
+2. Ship one excellent code-review/release-gate workflow that exports a PR-ready
+   Markdown audit and refuses approval on unresolved critical claims.
+3. Add session import/export and a stable graph resource in JSON, Mermaid, and
+   DOT forms.
+4. Move tool, engine, protocol, and state domains out of `main.rs` without
+   changing the single-crate or public MCP contracts.
+5. Benchmark core-pack tool-call efficiency and governance accuracy against a
+   no-MCP baseline and a scratchpad baseline using planted failures.
+
+Exit signal: a new host can complete a source-backed PR go/no-go flow and use
+the exported audit without hand-editing it.
+
+### v0.4: opt-in depth
+
+1. Bind a real retrieval provider for evidence lookup; never silently fall back
+   to empty or fabricated results.
+2. Add optional embedding-based diversity and model-backed expansion/scoring
+   with provider identity, cost, and provenance on every output.
+3. Add SQLite behind an optional feature for large histories and cross-session
+   failure-pattern analysis; retain JSON as the zero-dependency default.
+4. Prototype a read-only MCP Apps graph/audit viewer.
+5. Add streamable HTTP only after multi-user demand justifies authentication,
+   tenancy, rate-limit, and deployment complexity.
+
+### v1.0: stable governance protocol
+
+1. Freeze core tool contracts, resource URIs, error codes, and tool-pack SemVer
+   rules.
+2. Publish a host compatibility matrix and threat model covering prompt
+   injection, evidence spoofing, redaction, oversized payloads, and replay.
+3. Require migration fixtures for persisted sessions and independent release
+   certification for package, Registry, native binaries, and protocol behavior.
+
+## Measurement contract
+
+Track evidence rather than feature count:
+
+- median install-to-first-audit time;
+- tool calls required for the core decision path;
+- critical planted claims that remain `DATA_DEFICIT`;
+- unsafe consensus attempts rejected with `consensus_blocked`;
+- audits usable as PR/ADR artifacts without editing;
+- compatibility pass rate by host and MCP protocol revision;
+- package-to-source reproducibility and advisory-warning count.
+
+## Guardrails
+
+- No implicit server-side LLM.
+- No fabricated retrieval or semantic confidence.
+- No new paradigm merely to increase tool count.
+- No HTTP service before an authenticated multi-user requirement exists.
+- No live publication, Registry, or benchmark claim without current evidence.
diff --git a/docs/README.md b/docs/README.md
index 9d3688f..170ddaf 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -247,8 +247,15 @@ The Thinking Modes are supported by reusable layers:
 | Graph | Preserve dependencies, conflicts, refinements, and merges. | `link_thoughts`, `merge_thought_branches`, `reasoning://session/{id}/graph` |
 | Audit | Replay and export what happened. | `replay_reasoning_session`, `export_reasoning_audit`, `export_memory_snapshot` |
 | Mode Catalog | Discover modes, aliases, costs, tools, safety rules, and compatibility values. | `reasoning://thinking-modes` |
+| Tool Packs | Discover the active core/standard/full surface and compatibility rollback. | `reasoning://config/tool-packs` |
 | Aliases | Team-specific natural-language triggers. | `set_reasoning_aliases`, `reasoning://config/aliases` |
 
+The server defaults to the 13-tool `core` pack. Set
+`REASONKIT_TOOL_PACK=standard` for 33 everyday workflow tools or
+`REASONKIT_TOOL_PACK=full` for the complete 47-tool compatibility surface.
+Invalid values warn on stderr and fall back to core. Packs are selected at
+startup and require a server restart to change.
+
 ## Evidence Contract
 
 ReasonKit Think is intentionally conservative:
@@ -286,6 +293,10 @@ DATA_DEFICIT or SOURCE_CONFLICT where needed, and tell me whether we can proceed
 Core:
 
 - `ARCHITECTURE.md` - technical architecture, execution model, provider boundary.
+- `MCP_STACK_RESEARCH.md` - primary-source ecosystem findings, product wedge,
+  and the sequenced post-v0.2 opportunity map.
+- `adr/0001-progressive-disclosure-and-governance-integrity.md` - accepted v0.2
+  product and protocol decision.
 
 Primary tools:
 
diff --git a/docs/adr/0001-progressive-disclosure-and-governance-integrity.md b/docs/adr/0001-progressive-disclosure-and-governance-integrity.md
new file mode 100644
index 0000000..d2db9a3
--- /dev/null
+++ b/docs/adr/0001-progressive-disclosure-and-governance-integrity.md
@@ -0,0 +1,50 @@
+# ADR 0001: Progressive disclosure and governance integrity first
+
+- Status: Accepted
+- Date: 2026-08-23
+
+```yaml
+decision: >
+  Make ReasonKit Think the agent-native reasoning governance layer, with a small
+  complete default tool pack and evidence-gated decisions, before adding more
+  reasoning paradigms or an embedded model.
+context: >
+  The server already implements rich graph, pipeline, verification, assumption,
+  checkpoint, ReAct, prompt, and resource surfaces. Its 47-tool default is hard
+  for hosts to select correctly, while several integrity defects can let later
+  state erase or bypass earlier blockers. The current SDK pin also trails the
+  current MCP protocol implementation.
+alternatives:
+  - Keep all tools visible by default and improve descriptions only.
+  - Reduce the product to a sequential-thinking scratchpad.
+  - Add an implicit server-side model to produce semantic content.
+rationale: >
+  A complete 13-tool core retains route, deliberate, author, coach, assume,
+  verify, gate, decide, and audit operations while reducing discovery load by
+  72 percent. Fail-closed state transitions and reproducible evaluations turn
+  the governance claim into observable product behavior.
+risks:
+  - Existing clients may expect a tool outside the default pack.
+  - Stricter gates can expose workflows that previously passed incorrectly.
+  - An SDK migration can change wire shapes despite compiling cleanly.
+mitigations:
+  - Preserve REASONKIT_TOOL_PACK=full as the compatibility escape hatch.
+  - Return actionable structured errors and pack-aware guidance.
+  - Isolate the SDK migration and test every supported stdio protocol version.
+validation:
+  - Exact 13, 33, and 47 tool inventories are enforced by tests.
+  - Planted-failure evaluations must remain blocked.
+  - Current and legacy MCP initialization and tool calls pass smoke tests.
+  - Formatting, clippy, unit tests, dependency audit, and release build pass.
+confidence: 0.9
+```
+
+## Boundaries
+
+- The host model remains the semantic engine; the server coordinates, records,
+  verifies, and gates host-authored reasoning.
+- Stdio remains the default and only transport in this tranche.
+- Heuristic outputs remain labeled. No silent model, retrieval, or network
+  fallback is introduced.
+- New academic paradigms, a hosted service, and a standalone UI are deferred
+  until the adoption and governance contracts have evidence.
diff --git a/docs/clients/claude-code.md b/docs/clients/claude-code.md
index 968e2ad..e4b2155 100644
--- a/docs/clients/claude-code.md
+++ b/docs/clients/claude-code.md
@@ -2,35 +2,40 @@
 
 ## Primary references
 
-- https://platform.claude.com/docs/en/agents-and-tools/mcp-connector
-- https://modelcontextprotocol.io/docs/develop/connect-local-servers
+- <https://code.claude.com/docs/en/mcp>
+- <https://modelcontextprotocol.io/docs/develop/connect-local-servers>
 
-## Status
+## Register
 
-- Confidence: Medium-high (MCP connector + MCP standard docs).
+Install the release-candidate checkout on `PATH` as described in the root
+README, then register the stdio server:
 
-## Integration notes
-
-- Prefer local stdio server registration first.
-- Use debug mode if handshake fails.
-
-### Example (`~/.claude.json` → `mcpServers` entry)
+```sh
+claude mcp add --scope user reasonkit-think -- reasonkit-think-mcp
+claude mcp get reasonkit-think
+```
 
-Merge this object into the existing `mcpServers` map (do not duplicate the outer key):
+The equivalent `mcpServers` entry is:
 
 ```json
-"reasonkit-think": {
-  "type": "stdio",
-  "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
-  "args": [],
-  "env": { "TMPDIR": "/tmp" }
+{
+  "reasonkit-think": {
+    "type": "stdio",
+    "command": "reasonkit-think-mcp",
+    "args": []
+  }
 }
 ```
 
-Multi-client reference: root `README.md`.
+The default `core` pack exposes the 13-tool audited decision path. Set
+`REASONKIT_TOOL_PACK=full` on this server only as a compatibility escape hatch
+for a client that requires the complete 47-tool legacy inventory.
+
+## Verify
 
-## Verification checklist
+- The server initializes without text on stdout before the MCP handshake.
+- Claude discovers 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-- Server initializes successfully.
-- Tools and prompt are discovered.
-- Deliberation flow returns expected JSON outputs.
+If startup reports an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches Claude Code.
diff --git a/docs/clients/codex-cli.md b/docs/clients/codex-cli.md
index 25a3a4f..d0ea310 100644
--- a/docs/clients/codex-cli.md
+++ b/docs/clients/codex-cli.md
@@ -2,34 +2,36 @@
 
 ## Official references
 
-- https://developers.openai.com/codex/cli/reference
-- https://developers.openai.com/codex/config-reference
-- https://github.com/openai/codex/blob/main/docs/config.md
+- <https://developers.openai.com/codex/cli/reference>
+- <https://developers.openai.com/codex/config-reference>
 
-## Status
+## Register
 
-- Confidence: High (official OpenAI docs found).
+Install the release-candidate checkout on `PATH` as described in the root
+README, then register the stdio server:
 
-## Integration notes
-
-- Configure MCP server in Codex config.
-- Use stdio mode for local server process.
+```sh
+codex mcp add reasonkit-think -- reasonkit-think-mcp
+codex mcp get reasonkit-think
+```
 
-### Example (`~/.codex/config.toml`)
+The equivalent `~/.codex/config.toml` entry is:
 
 ```toml
 [mcp_servers.reasonkit-think]
-command = "/ABSOLUTE/PATH/TO/reasonkit-think-mcp"
+command = "reasonkit-think-mcp"
 args = []
-
-[mcp_servers.reasonkit-think.env]
-TMPDIR = "/tmp"
 ```
 
-Multi-client reference: root `README.md`.
+The default `core` pack exposes the 13-tool audited decision path. Set
+`REASONKIT_TOOL_PACK=full` on this server only as a compatibility escape hatch
+for a client that requires the complete 47-tool legacy inventory.
+
+## Verify
 
-## Verification checklist
+- `codex mcp get reasonkit-think` reports the server as enabled.
+- Tool discovery returns exactly 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-- Codex sees server
-- `tools/list` returns v1 + v2 tools
-- Deliberation commands execute without schema errors
+If startup reports an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches Codex.
diff --git a/docs/clients/copilot-vscode.md b/docs/clients/copilot-vscode.md
index 2120ffc..30a9716 100644
--- a/docs/clients/copilot-vscode.md
+++ b/docs/clients/copilot-vscode.md
@@ -1,74 +1,61 @@
-# Client Setup: Copilot CLI and VS Code (including Insiders)
+# Client Setup: Copilot CLI and VS Code
 
 ## Official references
 
-- https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-mcp-servers
-- https://docs.github.com/en/copilot/concepts/context/mcp
-- https://code.visualstudio.com/docs/copilot/customization/mcp-servers
-- https://code.visualstudio.com/docs/copilot/reference/mcp-configuration
+- <https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-mcp-servers>
+- <https://docs.github.com/en/copilot/concepts/context/mcp>
+- <https://code.visualstudio.com/docs/copilot/customization/mcp-servers>
+- <https://code.visualstudio.com/docs/copilot/reference/mcp-configuration>
 
-## Status
+## Register in Copilot CLI
 
-- Confidence: High (official GitHub and VS Code docs found).
+Install the release-candidate checkout on `PATH` as described in the root
+README, then register the local server:
 
-## Integration notes
-
-- Configure local stdio MCP server in Copilot CLI and VS Code MCP settings.
-- Ensure workspace trust before enabling server.
-- Use the absolute path to the built release binary, usually
-  `/path/to/reasonkit-think/target/release/reasonkit-think-mcp`.
+```sh
+copilot mcp add reasonkit-think -- reasonkit-think-mcp
+copilot mcp get reasonkit-think
+```
 
-### Copilot CLI example (`~/.copilot/mcp-config.json` or workspace `.mcp.json`)
+The equivalent `.mcp.json` entry is:
 
 ```json
 {
   "mcpServers": {
     "reasonkit-think": {
       "type": "local",
-      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
+      "command": "reasonkit-think-mcp",
       "args": [],
-      "env": { "TMPDIR": "/tmp" },
       "tools": ["*"]
     }
   }
 }
 ```
 
-### VS Code / VS Code Insiders example (`mcp.json`)
+## Register in VS Code
 
-VS Code stores MCP servers under the top-level `servers` key.
+VS Code stores MCP servers under the top-level `servers` key:
 
 ```json
 {
   "servers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
-      "args": [],
-      "env": { "TMPDIR": "/tmp" }
+      "command": "reasonkit-think-mcp",
+      "args": []
     }
   }
 }
 ```
 
-Multi-client reference: root `README.md`.
+The default `core` pack exposes exactly 13 tools. Set
+`REASONKIT_TOOL_PACK=full` only as the 47-tool compatibility escape hatch.
 
-## Verification checklist
+## Verify
 
-- server added and trusted
-- tools visible in MCP panel
-- prompts visible in MCP panel
-- reasoning resources accessible through `list_reasoning_resources` and `read_reasoning_resource`
-- `./scripts/client_compat_check.py` reports `status: "ok"`
+- The server is trusted and connected.
+- Tool discovery returns exactly 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-## Troubleshooting
-
-If Copilot reports `spawn ... ENOENT` or `Failed to connect to MCP server "reasonkit-think"`, the configured binary path does not exist or is not executable. Rebuild and verify:
-
-```bash
-cd /path/to/reasonkit-think
-export CARGO_TARGET_DIR="$PWD/target"
-cargo build --release
-./scripts/client_compat_check.py
-copilot mcp get reasonkit-think
-```
+If startup reports `ENOENT` or an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches the client.
diff --git a/docs/clients/cursor.md b/docs/clients/cursor.md
index aab5288..fde72ef 100644
--- a/docs/clients/cursor.md
+++ b/docs/clients/cursor.md
@@ -2,37 +2,34 @@
 
 ## Official references
 
-- https://cursor.com/docs/mcp
-- https://cursor.com/docs/cli/mcp
+- <https://docs.cursor.com/context/model-context-protocol>
+- <https://docs.cursor.com/cli/mcp>
 
-## Status
+## Register
 
-- Confidence: High (official docs found).
-
-## Integration notes
-
-- Configure server command to run `reasonkit-think-mcp` over stdio.
-- Validate `tools/list` and `prompts/list` immediately after registration.
-
-### Example (`~/.cursor/mcp.json`)
+Install the release-candidate checkout on `PATH` as described in the root
+README, then add this entry to `~/.cursor/mcp.json`:
 
 ```json
 {
   "mcpServers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
-      "args": [],
-      "env": { "TMPDIR": "/tmp" }
+      "command": "reasonkit-think-mcp",
+      "args": []
     }
   }
 }
 ```
 
-Full multi-client walkthrough: root `README.md` (collapsible sections).
+The default `core` pack exposes exactly 13 tools. Set
+`REASONKIT_TOOL_PACK=full` only as the 47-tool compatibility escape hatch.
+
+## Verify
 
-## Verification checklist
+- The MCP server appears as connected in Cursor's integrations view.
+- Tool discovery returns exactly 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-- MCP server appears in Cursor integrations list.
-- `sequentialthinking_tools` callable.
-- `sequential-thinking-guidance` visible.
+If startup reports an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches Cursor.
diff --git a/docs/clients/gemini-cli.md b/docs/clients/gemini-cli.md
index 74576c3..2cdf852 100644
--- a/docs/clients/gemini-cli.md
+++ b/docs/clients/gemini-cli.md
@@ -2,38 +2,41 @@
 
 ## Official references
 
-- https://geminicli.com/docs/
-- https://geminicli.com/docs/tools/mcp-server/
-- https://github.com/google-gemini/gemini-cli
+- <https://geminicli.com/docs/tools/mcp-server/>
+- <https://github.com/google-gemini/gemini-cli>
 
-## Status
+## Register
 
-- Confidence: High (official docs and repo references).
+Install the release-candidate checkout on `PATH` as described in the root
+README, then register the stdio server:
 
-## Integration notes
-
-- Register as stdio MCP server.
-- Validate via Gemini CLI MCP diagnostics commands.
-- Function declaration schemas are published in a Vertex/Gemini-friendly shape:
-  optional parameters use an explicit `type` plus `nullable: true`, not nullable
-  `anyOf` wrappers. If diagnostics fail, inspect `tools/list` and confirm
-  `consensus_answer.method` is a nullable string enum.
+```sh
+gemini mcp add --scope user reasonkit-think reasonkit-think-mcp
+gemini mcp list
+```
 
-### Example (`~/.gemini/settings.json` → `mcpServers`)
+The equivalent `mcpServers` entry is:
 
 ```json
-"reasonkit-think": {
-  "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
-  "args": [],
-  "env": { "TMPDIR": "/tmp" },
-  "trust": true
+{
+  "reasonkit-think": {
+    "command": "reasonkit-think-mcp",
+    "args": [],
+    "trust": true
+  }
 }
 ```
 
-Multi-client reference: root `README.md`.
+The server emits Gemini-compatible tool schemas: optional parameters use an
+explicit type plus `nullable: true`. The default `core` pack exposes exactly 13
+tools; use `REASONKIT_TOOL_PACK=full` only as the 47-tool compatibility escape
+hatch.
+
+## Verify
 
-## Verification checklist
+- `gemini mcp list` shows a connected local server.
+- Tool discovery returns exactly 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-- server listed
-- tools callable
-- resources retrievable
+If startup reports an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches Gemini CLI.
diff --git a/docs/clients/opencode.md b/docs/clients/opencode.md
index cc79d83..5e102ee 100644
--- a/docs/clients/opencode.md
+++ b/docs/clients/opencode.md
@@ -1,43 +1,41 @@
 # Client Setup: OpenCode
 
-## References
+## Official references
 
-- https://opencode.ai/docs/config/
-- https://opencode.ai/docs/tools/
-- https://opencode.ai/docs/server/
+- <https://opencode.ai/v2/docs/config>
+- <https://opencode.ai/v2/docs/mcp-servers>
 
-## Status
+## Register
 
-- Confidence: Medium (official docs found, verify MCP option details against current version before release).
-
-## Integration notes
-
-- Add MCP server config under OpenCode config schema (`mcp.<name>` with `"type": "local"` and a `command` array).
-
-### Example (`~/.opencode/opencode.json` or project `opencode.json`)
+Install the release-candidate checkout on `PATH` as described in the root
+README, then add this entry to `~/.config/opencode/opencode.json` or a project
+`opencode.json`:
 
 ```json
 {
   "$schema": "https://opencode.ai/config.json",
   "mcp": {
-    "reasonkit-think": {
-      "type": "local",
-      "command": ["/ABSOLUTE/PATH/TO/reasonkit-think-mcp"],
-      "enabled": true,
-      "environment": { "TMPDIR": "/tmp" }
+    "servers": {
+      "reasonkit-think": {
+        "type": "local",
+        "command": ["reasonkit-think-mcp"]
+      }
     }
   }
 }
 ```
 
-- **Config paths (important):** OpenCode merges multiple layers (see [Config](https://opencode.ai/docs/config/)). A file named `opencode.json` in your **home directory** (`~/opencode.json`) is easy to confuse with the **user/global** file OpenCode tends to ship with installs: **`~/.opencode/opencode.json`**. If you only edit `~/opencode.json`, the UI may still load MCP from `~/.opencode/opencode.json` unless you point **`OPENCODE_CONFIG`** at your preferred file or duplicate entries there.
-- **Project repos:** Working inside a Git repo usually picks up **`opencode.json` in that project root** (highest precedence for standard file-based config). Add `reasonkit-think` under `mcp` there too when you develop in a repo that defines its own `mcp` block.
-- Validate tool invocation and output rendering.
+OpenCode merges configuration layers; a project file can override the user
+entry. OpenCode v2 nests named servers under `mcp.servers`; use `disabled: true`
+only when intentionally turning one off. The default `core` pack exposes
+exactly 13 tools. Set
+`REASONKIT_TOOL_PACK=full` only as the 47-tool compatibility escape hatch.
 
-Multi-client reference: root `README.md`.
+## Verify
 
-## Verification checklist
+- The local server starts without stdout noise before the MCP handshake.
+- Tool discovery returns exactly 13 tools under the default pack.
+- The Auto → evidence → checkpoint/decision → audit path completes.
 
-- server starts
-- tools list loads
-- prompt/resource fetch works
+If startup reports an executable error, run
+`command -v reasonkit-think-mcp` in the environment that launches OpenCode.
diff --git a/docs/dependency-audit.md b/docs/dependency-audit.md
new file mode 100644
index 0000000..1b8ccc3
--- /dev/null
+++ b/docs/dependency-audit.md
@@ -0,0 +1,15 @@
+# Dependency audit policy
+
+CI runs `cargo audit --deny warnings`. Vulnerabilities, unmaintained packages,
+unsound advisories, and yanked dependencies all fail the gate.
+
+The v0.2 dependency cleanup removed the dormant `reasonkit-mem` integration.
+Production never populated its hot cache, while the dependency pulled 199
+additional locked packages and all eight prior audit warnings. The
+compatibility-only `think_query` tool remains discoverable in the full pack but
+now returns `experimental_memory_unavailable`; it performs no retrieval and
+cannot create a successful-looking empty result.
+
+Current lockfile expectation: `cargo audit --deny warnings` exits successfully
+with zero warnings. A future memory provider must restore retrieval only with a
+real population path, provenance, contract tests, and a warning-clean graph.
diff --git a/docs/plans/2026-08-23-next-level-v0-2.md b/docs/plans/2026-08-23-next-level-v0-2.md
new file mode 100644
index 0000000..8b93ea2
--- /dev/null
+++ b/docs/plans/2026-08-23-next-level-v0-2.md
@@ -0,0 +1,59 @@
+# ReasonKit Think v0.2 implementation plan
+
+## Goal
+
+Deliver a materially easier and more trustworthy OSS MCP server: a complete
+five-minute path from intent to auditable decision, current protocol support,
+fail-closed state transitions, repeatable evaluations, and honest release and
+discovery metadata.
+
+## Workstream 1: MCP protocol and progressive disclosure
+
+1. Upgrade `rmcp` to the current compatible SDK in an isolated commit.
+2. Preserve stdio and the existing capability boundary.
+3. Add `REASONKIT_TOOL_PACK=core|standard|full`, defaulting to the audited
+   13-tool core; preserve exact 33-tool standard and 47-tool full inventories.
+4. Store a filtered tool router and expose `reasoning://config/tool-packs`.
+5. Make startup instructions and guidance pack-aware.
+6. Return both JSON text and MCP `structuredContent` from common result helpers.
+7. Test current and legacy protocol negotiation and exact discovery inventories.
+
+## Workstream 2: Governance integrity
+
+1. Upsert verification claims without erasing unrelated prior blockers.
+2. Require qualifying evidence before an assumption can become verified.
+3. Accept observations only for action nodes and complete the referenced action.
+4. Reject unknown pipeline stages rather than synthesizing a passing stage.
+5. Generate one audit identifier and reuse it in the envelope and payload.
+6. Add regression tests for every corrected bypass.
+
+## Workstream 3: Adoption, evidence, and reproducibility
+
+1. Rewrite the README golden path around Auto mode, verification, decision, and
+   audit; move specialist surfaces behind explicit power-user guidance.
+2. Add deterministic planted-failure evaluations with machine-readable metrics.
+3. Run tests and evaluations in CI, and publish truthful package/registry metadata.
+4. Add release automation only when its generated configuration validates.
+5. Quarantine incomplete memory/retrieval behavior behind an explicit
+   experimental feature or remove claims that exceed implemented behavior.
+6. Reconcile roadmap files so shipped, partial, and proposed work are distinct.
+
+## Integration and verification
+
+1. Land each workstream as independently reviewable commits from isolated
+   worktrees.
+2. Cross-review the merged diff for protocol compatibility, trust regressions,
+   security, and OSS-boundary violations.
+3. Run formatting, clippy with warnings denied, all tests, doctests, release
+   build, MCP smoke tests, planted-failure evaluations, and `cargo audit`.
+4. Verify that the user's original dirty checkout is unchanged.
+
+## Exit criteria
+
+- Core discovery exposes only the complete 13-tool golden path.
+- `full` restores all existing tools without data migration.
+- A critical verification blocker cannot be erased by a later verification call.
+- Unsupported stages and evidence-free verification fail closed.
+- Planted false-claim and unresolved-assumption evaluations do not proceed.
+- The release candidate passes every available local gate; any unavailable gate
+  is reported with its exact blocker.
diff --git a/docs/resources/config-providers.md b/docs/resources/config-providers.md
index ef4dda1..b0da320 100644
--- a/docs/resources/config-providers.md
+++ b/docs/resources/config-providers.md
@@ -18,3 +18,10 @@ boundaries.
 
 Configured endpoint URLs are not returned. The resource reports only whether an
 endpoint is present.
+
+## Doctor contract
+
+`reasonkit-think-mcp doctor --json` reports an unset or `agent` `BACKEND_MODE`
+as healthy. Invalid values and `external` report a warning and an effective
+`backend_mode` of `agent`; `external_provider_bound` remains `false` until a
+production provider is bound. Diagnostics never interpolate an invalid value.
diff --git a/docs/resources/tool-packs.md b/docs/resources/tool-packs.md
new file mode 100644
index 0000000..90d9d42
--- /dev/null
+++ b/docs/resources/tool-packs.md
@@ -0,0 +1,24 @@
+# Tool Packs
+
+Resource URI: `reasoning://config/tool-packs`
+
+ReasonKit Think selects one immutable tool router at process startup:
+
+| Pack | Count | Purpose |
+| --- | ---: | --- |
+| `core` | 13 | Default agent-native authoring and fail-closed governance path. |
+| `standard` | 33 | Core plus everyday Explore, Map, Sketch, replay, tuning, and resource helpers. |
+| `full` | 47 | Complete compatibility surface, including legacy and specialist integrations. |
+
+Set `REASONKIT_TOOL_PACK` in the MCP server environment. Unset or empty means
+`core`. An invalid value writes a warning to stderr and uses the safe core
+fallback. Restart the stdio process after changing it.
+
+Use `REASONKIT_TOOL_PACK=full` to restore the complete pre-pack tool surface.
+The resource returns the active/default pack, exact ordered membership of every
+pack, counts, restart requirement, and compatibility rollback instruction. It
+never returns environment values or secrets.
+
+The `full` pack retains the experimental `think_query` name for client contract
+compatibility. Until a production memory ingestion and retrieval provider is
+bound, it returns `experimental_memory_unavailable` and performs no retrieval.
diff --git a/docs/tools/advanced_reasoning_operators.md b/docs/tools/advanced_reasoning_operators.md
index 880348a..b7e3535 100644
--- a/docs/tools/advanced_reasoning_operators.md
+++ b/docs/tools/advanced_reasoning_operators.md
@@ -9,6 +9,11 @@ binding supplies real model-backed or retrieval-backed execution.
 - `record_assumption` stores a premise with status, criticality, optional
   evidence, and source node links.
 - `set_assumption_status` updates that premise after evidence is gathered.
+- `critical: true` is authoritative when legacy `critical` and `criticality`
+  fields conflict; callers cannot downgrade a blocker by also sending a lower
+  criticality.
+- A requested `verified` state remains `unresolved` unless the supplied
+  evidence satisfies the active verification policy.
 - Critical/high verifiable assumptions that are not `verified` block
   `consensus_answer`, `converge_reasoning`, and checkpoints under fail-closed
   policy.
diff --git a/docs/tools/consensus_answer.md b/docs/tools/consensus_answer.md
index 4a7afb4..5dc28c3 100644
--- a/docs/tools/consensus_answer.md
+++ b/docs/tools/consensus_answer.md
@@ -19,7 +19,7 @@ semantic synthesis or calibrated confidence.
   - `require_tier1_unambiguous_for_critical`: boolean
   - `fail_closed_on_critical_unresolved`: boolean
 
-## Output
+## Successful output
 
 - `final_answer`
 - `confidence`
@@ -29,15 +29,30 @@ semantic synthesis or calibrated confidence.
 - `disagreement_report`
 - `analysis`: provenance and limitations for the consensus run
 
+When governance blocks convergence, the MCP tool returns an application error
+with `code: "consensus_blocked"`, the safe route, blocker summary, and a
+recommended next action. That blocker response deliberately omits
+`final_answer`; clients must not present an unapproved candidate answer as a
+decision.
+
 ## Behavior
 
-- Fails closed on unresolved critical verification issues.
-- Requires route decision in governed mode.
-- Returns `policy_blocked = true` when fail-closed policy blocks the consensus result.
+- Fails closed on unresolved critical verification issues or assumptions.
+- Requires at least one agent-authored frontier candidate and excludes
+  heuristic scaffold/root nodes from an allowed final answer.
+- Authorship requires a Thought node with server-recorded `agent_authored`
+  provenance; a caller-added tag on an Action or other heuristic node cannot
+  satisfy the gate. Consensus provenance lists only the candidates actually
+  used in the answer.
+- Blocks when a pipeline stage failed its gate or the latest checkpoint still
+  reports blocking gaps.
+- Requires a route decision in governed mode.
+- A `policy_override` applies only to that call; it does not silently rewrite
+  the session policy.
 - `confidence` is a route heuristic derived from verification status, not a
   model self-rating or historical calibration score.
 
 ## Errors
 
 - No eligible branches
-- Policy gate failure
+- Policy, authorship, pipeline, or checkpoint gate failure
diff --git a/docs/tools/export_reasoning_audit.md b/docs/tools/export_reasoning_audit.md
index 3509ea0..e14414b 100644
--- a/docs/tools/export_reasoning_audit.md
+++ b/docs/tools/export_reasoning_audit.md
@@ -2,23 +2,33 @@
 
 ## Purpose
 
-Export a complete machine-readable reasoning artifact for downstream agents and compliance review.
+Export the canonical reasoning payload plus a deterministic machine- or human-readable artifact for downstream agents and compliance review.
 
 ## Input
 
-- `deliberation_id`
-- `format`: `json`
-- `include_raw_thoughts`: bool
+- `deliberation_id` (required)
+- `format` (optional): `json` (default) or `markdown`
+- `include_raw_thoughts` (optional bool, default `false`)
 
 ## Output
 
+- `deliberation_id`
 - `audit_id`
-- `export_path` or `payload`
+- `payload`: the unchanged canonical JSON audit object
+- `format`: the selected format
+- `content_type`: `application/json` or `text/markdown; charset=utf-8`
+- `artifact`: deterministic rendering of the canonical audit snapshot
 
 ## Includes
 
-- session metadata
-- graph snapshot
-- verification matrix
-- final route decision
-- confidence and unresolved risks
+- session, mode, goal, and route metadata
+- verification statuses and evidence sources
+- assumptions, checkpoints, pipeline gates, and blockers
+- explicit heuristic/provider provenance
+- graph metadata; raw node content only when `include_raw_thoughts=true`
+
+Markdown tables escape pipes, line breaks, HTML-significant characters, and
+active Markdown delimiters for links, images, emphasis, code, and strikethrough.
+Map-derived sections are sorted for stable output. JSON remains the
+compatibility default, and the canonical `payload`, `audit_id`, and
+`deliberation_id` fields remain available in both formats.
diff --git a/docs/tools/run_thinking_mode.md b/docs/tools/run_thinking_mode.md
index e3133e7..841d6d2 100644
--- a/docs/tools/run_thinking_mode.md
+++ b/docs/tools/run_thinking_mode.md
@@ -86,6 +86,11 @@ The result always includes:
 - `policy_blocked`: whether unresolved critical evidence blocked confidence.
 - `audit_id`: exported audit artifact id.
 
+The returned graph can contain heuristic scaffolds. Follow `next_action` and
+submit the host's actual analysis through `add_thought_node` before requesting
+an allowed `consensus_answer`; a root goal or scaffold is never a consensus
+candidate.
+
 Mode-specific fields appear under `quick`, `explore`, `map`, `sketch`, or `test`.
 
 ## Behavior Contract
diff --git a/evals/README.md b/evals/README.md
new file mode 100644
index 0000000..6f325cb
--- /dev/null
+++ b/evals/README.md
@@ -0,0 +1,28 @@
+# Contract evaluations
+
+`tool-packs.json` is the exact discovery contract for the lexical `core` (13),
+`standard` (33), and `full` (47) tool packs. `run_contract_evals.py` starts a
+fresh stdio server for every tool-pack/protocol pair, validates typed schemas,
+and then enforces eight adversarial governance checks: evidence-free assumption
+verification, conflicting assumption criticality, normalized-claim criticality
+downgrades, checkpoint routing, blocker-only consensus errors, unknown pipeline
+stages, call-scoped policy overrides, and scaffold-only convergence.
+
+Run the deterministic unit layer first:
+
+```sh
+just unit
+```
+
+Run the live release-binary matrix after the protocol and governance changes
+are integrated:
+
+```sh
+just eval
+```
+
+The live matrix covers `2024-11-05`, `2025-03-26`, `2025-06-18`,
+`2025-11-25`, and `2026-07-28`. It is intentionally a cross-stream gate: a
+baseline binary without `REASONKIT_TOOL_PACK` or the fail-closed governance
+repairs will fail. Use `--surface-only` only for focused discovery diagnosis;
+CI runs the complete evaluation.
diff --git a/evals/__init__.py b/evals/__init__.py
new file mode 100644
index 0000000..a9493a2
--- /dev/null
+++ b/evals/__init__.py
@@ -0,0 +1 @@
+"""Deterministic adoption and governance evaluations for ReasonKit Think."""
diff --git a/evals/mcp_contract.py b/evals/mcp_contract.py
new file mode 100644
index 0000000..23000d3
--- /dev/null
+++ b/evals/mcp_contract.py
@@ -0,0 +1,291 @@
+from __future__ import annotations
+
+import json
+from pathlib import Path
+from typing import Any
+
+
+CONTRACT_PATH = Path(__file__).with_name("tool-packs.json")
+EXPECTED_PACK_COUNTS = {"core": 13, "standard": 33, "full": 47}
+
+
+def load_contract(path: Path | None = None) -> dict[str, Any]:
+    return json.loads((path or CONTRACT_PATH).read_text(encoding="utf-8"))
+
+
+def validate_contract(contract: dict[str, Any]) -> list[str]:
+    issues: list[str] = []
+    packs = contract.get("toolPacks")
+    if not isinstance(packs, dict):
+        return ["toolPacks must be an object"]
+
+    if contract.get("schemaVersion") != 1:
+        issues.append("schemaVersion must be 1")
+    if contract.get("defaultToolPack") != "core":
+        issues.append("defaultToolPack must be core")
+    if set(packs) != set(EXPECTED_PACK_COUNTS):
+        issues.append("toolPacks must contain exactly core, standard, and full")
+
+    for name, expected_count in EXPECTED_PACK_COUNTS.items():
+        tools = packs.get(name)
+        if not isinstance(tools, list) or not all(isinstance(tool, str) for tool in tools):
+            issues.append(f"{name} must be a list of tool names")
+            continue
+        if len(tools) != expected_count:
+            issues.append(f"{name} must contain exactly {expected_count} tools")
+        if tools != sorted(tools):
+            issues.append(f"{name} must use lexical router order")
+        if len(tools) != len(set(tools)):
+            issues.append(f"{name} contains duplicate tools")
+
+    if all(isinstance(packs.get(name), list) for name in EXPECTED_PACK_COUNTS):
+        if not set(packs["core"]).issubset(packs["standard"]):
+            issues.append("core must be a subset of standard")
+        if not set(packs["standard"]).issubset(packs["full"]):
+            issues.append("standard must be a subset of full")
+
+    versions = contract.get("protocolVersions")
+    if not isinstance(versions, list) or len(versions) != len(set(versions or [])):
+        issues.append("protocolVersions must be a unique list")
+    return issues
+
+
+def validate_tool_surface(
+    contract: dict[str, Any], pack: str, tools: list[dict[str, Any] | str]
+) -> list[str]:
+    expected = contract["toolPacks"][pack]
+    actual = [tool if isinstance(tool, str) else tool.get("name") for tool in tools]
+    actual_names = [name for name in actual if isinstance(name, str)]
+    issues: list[str] = []
+
+    missing = sorted(set(expected) - set(actual_names))
+    unexpected = sorted(set(actual_names) - set(expected))
+    if missing:
+        issues.append(f"{pack} missing tools: {', '.join(missing)}")
+    if unexpected:
+        issues.append(f"{pack} unexpected tools: {', '.join(unexpected)}")
+    if actual_names != sorted(actual_names):
+        issues.append(f"{pack} tool list is not in lexical router order")
+    if not missing and not unexpected and actual_names != expected:
+        issues.append(f"{pack} tool order differs from the exact contract")
+    return issues
+
+
+def schema_type_issues(schema: object, path: str = "$") -> list[str]:
+    issues: list[str] = []
+    if not isinstance(schema, dict):
+        return issues
+
+    requires_type = any(
+        key in schema
+        for key in (
+            "properties",
+            "additionalProperties",
+            "items",
+            "anyOf",
+            "oneOf",
+            "allOf",
+            "enum",
+            "const",
+        )
+    )
+    if requires_type and "type" not in schema and "$ref" not in schema:
+        issues.append(f"{path}: missing type")
+
+    for key, child in schema.items():
+        if isinstance(child, dict):
+            issues.extend(schema_type_issues(child, f"{path}.{key}"))
+        elif isinstance(child, list):
+            for index, item in enumerate(child):
+                issues.extend(schema_type_issues(item, f"{path}.{key}[{index}]"))
+    return issues
+
+
+def tool_result_is_error(response: dict[str, Any]) -> bool:
+    result = response.get("result")
+    if not isinstance(result, dict):
+        return "error" in response
+    return result.get("isError") is True or result.get("is_error") is True
+
+
+def tool_payload(response: dict[str, Any]) -> dict[str, Any]:
+    result = response.get("result", {})
+    if not isinstance(result, dict):
+        return {}
+    structured = result.get("structuredContent") or result.get("structured_content")
+    if isinstance(structured, dict):
+        return structured
+    for item in result.get("content", []):
+        if isinstance(item, dict) and isinstance(item.get("text"), str):
+            try:
+                payload = json.loads(item["text"])
+            except json.JSONDecodeError:
+                continue
+            if isinstance(payload, dict):
+                return payload
+    return {}
+
+
+def contains_key(value: object, key: str) -> bool:
+    if isinstance(value, dict):
+        return key in value or any(contains_key(child, key) for child in value.values())
+    if isinstance(value, list):
+        return any(contains_key(child, key) for child in value)
+    return False
+
+
+def evaluate_fail_closed(
+    *,
+    evidence_free_update: dict[str, Any],
+    criticality_downgrade: dict[str, Any],
+    checkpoint: dict[str, Any],
+    consensus: dict[str, Any],
+    unknown_stage: dict[str, Any],
+    pipeline_override: dict[str, Any],
+    post_pipeline_audit: dict[str, Any],
+    scaffold_run: dict[str, Any],
+) -> dict[str, Any]:
+    update_payload = tool_payload(evidence_free_update)
+    update_assumption = update_payload.get("assumption", {})
+    update_blocked = (
+        isinstance(update_assumption, dict)
+        and update_assumption.get("status") != "verified"
+        and bool(update_payload.get("blocking_assumption_ids"))
+    )
+    assumption_criticality_preserved = (
+        isinstance(update_assumption, dict)
+        and update_assumption.get("criticality") == "critical"
+    )
+
+    checkpoint_payload = tool_payload(checkpoint).get("checkpoint", {})
+    checkpoint_blocked = (
+        isinstance(checkpoint_payload, dict)
+        and checkpoint_payload.get("route_decision")
+        in {"GATHER_MORE_EVIDENCE", "DEFER_TO_HUMAN", "REJECT_CURRENT_PLAN"}
+        and bool(checkpoint_payload.get("blocking_gaps"))
+    )
+
+    downgrade_matrix = tool_payload(criticality_downgrade).get(
+        "verification_matrix", []
+    )
+    criticality_preserved = isinstance(downgrade_matrix, list) and any(
+        isinstance(entry, dict)
+        and " ".join(str(entry.get("claim", "")).split()).casefold()
+        == "release requires approval"
+        and entry.get("critical") is True
+        and entry.get("status") == "data_deficit"
+        for entry in downgrade_matrix
+    )
+
+    consensus_payload = tool_payload(consensus)
+    consensus_blocked = (
+        tool_result_is_error(consensus)
+        and consensus_payload.get("code") == "consensus_blocked"
+        and not contains_key(consensus, "final_answer")
+    )
+    unknown_stage_rejected = (
+        tool_result_is_error(unknown_stage)
+        and tool_payload(unknown_stage).get("code") == "invalid_input"
+    )
+
+    audit_payload = tool_payload(post_pipeline_audit).get("payload", {})
+    policy = (
+        audit_payload.get("verification_policy", {})
+        if isinstance(audit_payload, dict)
+        else {}
+    )
+    pipeline_policy_call_scoped = (
+        not tool_result_is_error(pipeline_override)
+        and isinstance(policy, dict)
+        and policy.get("min_independent_groups") == 3
+        and policy.get("allow_tier3_for_independence") is False
+        and policy.get("require_tier1_unambiguous_for_critical") is True
+        and policy.get("fail_closed_on_critical_unresolved") is True
+    )
+
+    scaffold_payload = tool_payload(scaffold_run)
+    scaffold_authorship = scaffold_payload.get("graph_authorship", {})
+    scaffold_route_blocked = (
+        isinstance(scaffold_authorship, dict)
+        and scaffold_authorship.get("agent_authored") == 0
+        and scaffold_authorship.get("scaffold_nodes", 0) > 0
+        and scaffold_authorship.get("frontier_scaffold_only") is True
+        and scaffold_payload.get("policy_blocked") is True
+        and scaffold_payload.get("route_decision")
+        in {"GATHER_MORE_EVIDENCE", "DEFER_TO_HUMAN", "REJECT_CURRENT_PLAN"}
+        and scaffold_payload.get("pipeline_route")
+        in {"GATHER_MORE_EVIDENCE", "DEFER_TO_HUMAN", "REJECT_CURRENT_PLAN"}
+    )
+    checks = {
+        "evidence_free_verification_blocked": update_blocked,
+        "assumption_criticality_preserved": assumption_criticality_preserved,
+        "critical_claim_downgrade_blocked": criticality_preserved,
+        "critical_checkpoint_blocked": checkpoint_blocked,
+        "consensus_exact_blocker_only": consensus_blocked,
+        "unknown_pipeline_stage_invalid_input": unknown_stage_rejected,
+        "pipeline_policy_override_call_scoped": pipeline_policy_call_scoped,
+        "scaffold_only_route_blocked": scaffold_route_blocked,
+    }
+    return {"passed": all(checks.values()), "checks": checks}
+
+
+def matrix_jobs(
+    contract: dict[str, Any],
+    packs: list[str] | None = None,
+    protocols: list[str] | None = None,
+) -> list[tuple[str, str]]:
+    selected_packs = packs or list(contract["toolPacks"])
+    selected_protocols = protocols or contract["protocolVersions"]
+    unknown_packs = sorted(set(selected_packs) - set(contract["toolPacks"]))
+    unknown_protocols = sorted(
+        set(selected_protocols) - set(contract["protocolVersions"])
+    )
+    if unknown_packs or unknown_protocols:
+        raise ValueError(
+            f"unknown packs={unknown_packs!r} protocols={unknown_protocols!r}"
+        )
+    return [
+        (pack, protocol)
+        for pack in selected_packs
+        for protocol in selected_protocols
+    ]
+
+
+def validate_runtime_probe(
+    contract: dict[str, Any],
+    *,
+    pack: str,
+    protocol: str,
+    initialize: dict[str, Any],
+    tools: list[dict[str, Any]],
+) -> list[str]:
+    issues: list[str] = []
+    result = initialize.get("result")
+    if not isinstance(result, dict):
+        return [f"{pack}/{protocol}: initialization failed"]
+
+    server_info = result.get("serverInfo") or result.get("server_info")
+    server_name = server_info.get("name") if isinstance(server_info, dict) else None
+    if server_name != "reasonkit-think-mcp":
+        issues.append(
+            f"{pack}/{protocol}: server name is {server_name!r}, expected 'reasonkit-think-mcp'"
+        )
+
+    negotiated = result.get("protocolVersion") or result.get("protocol_version")
+    if negotiated != protocol:
+        issues.append(
+            f"{pack}/{protocol}: negotiated {negotiated!r}, expected {protocol!r}"
+        )
+
+    issues.extend(validate_tool_surface(contract, pack, tools))
+    for tool in tools:
+        name = tool.get("name", "<unnamed>")
+        schema = tool.get("inputSchema") or tool.get("input_schema")
+        if not isinstance(schema, dict):
+            issues.append(f"{pack}/{protocol}: {name} missing input schema")
+            continue
+        issues.extend(
+            f"{pack}/{protocol}: {name} {issue}"
+            for issue in schema_type_issues(schema)
+        )
+    return issues
diff --git a/evals/run_contract_evals.py b/evals/run_contract_evals.py
new file mode 100644
index 0000000..dbcdac4
--- /dev/null
+++ b/evals/run_contract_evals.py
@@ -0,0 +1,427 @@
+#!/usr/bin/env python3
+"""Run deterministic MCP discovery and fail-closed contract evaluations."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import select
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+from typing import Any
+
+if __package__ in (None, ""):
+    sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from evals.mcp_contract import (  # noqa: E402
+    evaluate_fail_closed,
+    load_contract,
+    matrix_jobs,
+    tool_payload,
+    validate_contract,
+    validate_runtime_probe,
+)
+
+
+class McpSession:
+    def __init__(self, binary: Path, pack: str, timeout_seconds: float) -> None:
+        self.binary = binary
+        self.pack = pack
+        self.timeout_seconds = timeout_seconds
+        self._request_id = 0
+        self._state_dir = tempfile.TemporaryDirectory(
+            prefix=f"reasonkit-think-eval-{pack}-"
+        )
+        env = os.environ.copy()
+        env.update(
+            {
+                "REASONKIT_TOOL_PACK": pack,
+                "RUST_LOG": "error",
+                "TMPDIR": "/tmp",
+                "XDG_DATA_HOME": str(Path(self._state_dir.name) / "data"),
+            }
+        )
+        self.proc = subprocess.Popen(
+            [str(binary)],
+            cwd="/",
+            env=env,
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            text=True,
+            bufsize=1,
+        )
+
+    def __enter__(self) -> McpSession:
+        return self
+
+    def __exit__(self, *_exc: object) -> None:
+        self.close()
+
+    def close(self) -> None:
+        if self.proc.poll() is None:
+            self.proc.terminate()
+            try:
+                self.proc.wait(timeout=2)
+            except subprocess.TimeoutExpired:
+                self.proc.kill()
+                self.proc.wait(timeout=2)
+        self._state_dir.cleanup()
+
+    def send(self, payload: dict[str, Any]) -> None:
+        if self.proc.stdin is None:
+            raise RuntimeError("MCP stdin is unavailable")
+        self.proc.stdin.write(json.dumps(payload, separators=(",", ":")) + "\n")
+        self.proc.stdin.flush()
+
+    def receive(self, request_id: int) -> dict[str, Any]:
+        if self.proc.stdout is None:
+            raise RuntimeError("MCP stdout is unavailable")
+        while True:
+            ready, _, _ = select.select(
+                [self.proc.stdout], [], [], self.timeout_seconds
+            )
+            if not ready:
+                raise TimeoutError(
+                    f"MCP response timed out for request {request_id} ({self.pack})"
+                )
+            line = self.proc.stdout.readline()
+            if not line:
+                stderr = self.proc.stderr.read() if self.proc.stderr else ""
+                raise RuntimeError(
+                    f"MCP server exited before request {request_id}: {stderr[-1000:]}"
+                )
+            response = json.loads(line)
+            if response.get("id") == request_id:
+                return response
+
+    def request(self, method: str, params: dict[str, Any]) -> dict[str, Any]:
+        self._request_id += 1
+        request_id = self._request_id
+        self.send(
+            {
+                "jsonrpc": "2.0",
+                "id": request_id,
+                "method": method,
+                "params": params,
+            }
+        )
+        return self.receive(request_id)
+
+    def initialize(self, protocol: str) -> dict[str, Any]:
+        response = self.request(
+            "initialize",
+            {
+                "protocolVersion": protocol,
+                "capabilities": {},
+                "clientInfo": {
+                    "name": "reasonkit-contract-eval",
+                    "version": "1.0.0",
+                },
+            },
+        )
+        if "result" in response:
+            self.send(
+                {
+                    "jsonrpc": "2.0",
+                    "method": "notifications/initialized",
+                    "params": {},
+                }
+            )
+        return response
+
+    def call_tool(self, name: str, arguments: dict[str, Any]) -> dict[str, Any]:
+        return self.request("tools/call", {"name": name, "arguments": arguments})
+
+
+def probe_surface(
+    binary: Path,
+    contract: dict[str, Any],
+    pack: str,
+    protocol: str,
+    timeout_seconds: float,
+) -> dict[str, Any]:
+    try:
+        with McpSession(binary, pack, timeout_seconds) as session:
+            initialize = session.initialize(protocol)
+            tools: list[dict[str, Any]] = []
+            if "result" in initialize:
+                listed = session.request("tools/list", {})
+                candidate = listed.get("result", {}).get("tools", [])
+                if isinstance(candidate, list):
+                    tools = [tool for tool in candidate if isinstance(tool, dict)]
+            issues = validate_runtime_probe(
+                contract,
+                pack=pack,
+                protocol=protocol,
+                initialize=initialize,
+                tools=tools,
+            )
+            return {
+                "pack": pack,
+                "protocol": protocol,
+                "toolCount": len(tools),
+                "status": "passed" if not issues else "failed",
+                "issues": issues,
+            }
+    except (OSError, RuntimeError, TimeoutError, json.JSONDecodeError) as error:
+        return {
+            "pack": pack,
+            "protocol": protocol,
+            "toolCount": 0,
+            "status": "failed",
+            "issues": [str(error)],
+        }
+
+
+def run_fail_closed_eval(
+    binary: Path, protocol: str, timeout_seconds: float
+) -> dict[str, Any]:
+    try:
+        with McpSession(binary, "core", timeout_seconds) as session:
+            initialized = session.initialize(protocol)
+            if "result" not in initialized:
+                return {
+                    "status": "failed",
+                    "passed": False,
+                    "checks": {},
+                    "issues": ["fail-closed session initialization failed"],
+                }
+
+            started = session.call_tool(
+                "start_deliberation",
+                {
+                    "session_id": "contract-eval-fail-closed",
+                    "mode": "reasonkit",
+                    "goal": "Reject an unsupported critical deployment claim",
+                },
+            )
+            deliberation_id = tool_payload(started).get("deliberation_id")
+            if not isinstance(deliberation_id, str):
+                return {
+                    "status": "failed",
+                    "passed": False,
+                    "checks": {},
+                    "issues": ["start_deliberation returned no deliberation_id"],
+                }
+
+            assumption = session.call_tool(
+                "record_assumption",
+                {
+                    "deliberation_id": deliberation_id,
+                    "text": "The deployment is safe despite having no supporting evidence",
+                    "criticality": "medium",
+                    "critical": True,
+                    "status": "verified",
+                    "verifiable": True,
+                },
+            )
+            assumption_id = (
+                tool_payload(assumption).get("assumption", {}).get("assumption_id")
+            )
+            if not isinstance(assumption_id, str):
+                return {
+                    "status": "failed",
+                    "passed": False,
+                    "checks": {},
+                    "issues": ["record_assumption returned no assumption_id"],
+                }
+
+            evidence_free_update = session.call_tool(
+                "set_assumption_status",
+                {
+                    "deliberation_id": deliberation_id,
+                    "assumption_id": assumption_id,
+                    "status": "verified",
+                    "confidence": 0.99,
+                    "notes": "planted evidence-free verification attempt",
+                },
+            )
+            session.call_tool(
+                "verify_thoughts",
+                {
+                    "deliberation_id": deliberation_id,
+                    "node_ids": [],
+                    "critical_claims": [],
+                    "claims": [
+                        {
+                            "text": "Release requires approval",
+                            "critical": True,
+                            "evidence": [],
+                        }
+                    ],
+                    "method": "hybrid",
+                },
+            )
+            criticality_downgrade = session.call_tool(
+                "verify_thoughts",
+                {
+                    "deliberation_id": deliberation_id,
+                    "node_ids": [],
+                    "critical_claims": [],
+                    "claims": [
+                        {
+                            "text": "  Release   requires approval  ",
+                            "critical": False,
+                            "evidence": [
+                                {
+                                    "source": f"secondary-source-{index}",
+                                    "tier": "tier2",
+                                    "independence_group": f"secondary-group-{index}",
+                                    "supports": True,
+                                    "contradictory": False,
+                                    "unambiguous": True,
+                                }
+                                for index in range(1, 4)
+                            ],
+                        }
+                    ],
+                    "method": "hybrid",
+                },
+            )
+            unknown_stage = session.call_tool(
+                "run_reasonkit_pipeline",
+                {
+                    "deliberation_id": deliberation_id,
+                    "stages": ["planted-unknown-stage"],
+                    "policy_override": "relaxed",
+                },
+            )
+            pipeline_override = session.call_tool(
+                "run_reasonkit_pipeline",
+                {
+                    "deliberation_id": deliberation_id,
+                    "stages": ["causal-modeling"],
+                    "policy_override": "relaxed",
+                },
+            )
+            post_pipeline_audit = session.call_tool(
+                "export_reasoning_audit",
+                {
+                    "deliberation_id": deliberation_id,
+                    "format": "json",
+                    "include_raw_thoughts": False,
+                },
+            )
+            checkpoint = session.call_tool(
+                "run_reasoning_checkpoint",
+                {
+                    "deliberation_id": deliberation_id,
+                    "label": "planted critical blocker",
+                    "fail_threshold": 0.0,
+                },
+            )
+            consensus = session.call_tool(
+                "consensus_answer", {"deliberation_id": deliberation_id}
+            )
+            scaffold_run = session.call_tool(
+                "run_thinking_mode",
+                {
+                    "session_id": "contract-eval-scaffold-only",
+                    "intent": "Compare several release strategies without host-authored analysis",
+                    "mode": "Explore",
+                },
+            )
+            evaluation = evaluate_fail_closed(
+                evidence_free_update=evidence_free_update,
+                criticality_downgrade=criticality_downgrade,
+                checkpoint=checkpoint,
+                consensus=consensus,
+                unknown_stage=unknown_stage,
+                pipeline_override=pipeline_override,
+                post_pipeline_audit=post_pipeline_audit,
+                scaffold_run=scaffold_run,
+            )
+            return {
+                "status": "passed" if evaluation["passed"] else "failed",
+                **evaluation,
+                "issues": [] if evaluation["passed"] else ["fail-closed bypass detected"],
+            }
+    except (OSError, RuntimeError, TimeoutError, json.JSONDecodeError) as error:
+        return {
+            "status": "failed",
+            "passed": False,
+            "checks": {},
+            "issues": [str(error)],
+        }
+
+
+def parse_args() -> argparse.Namespace:
+    contract = load_contract()
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--binary",
+        type=Path,
+        default=Path(os.environ.get("REASONKIT_THINK_BIN", "target/release/reasonkit-think-mcp")),
+    )
+    parser.add_argument(
+        "--packs",
+        nargs="+",
+        choices=list(contract["toolPacks"]),
+        default=list(contract["toolPacks"]),
+    )
+    parser.add_argument(
+        "--protocols",
+        nargs="+",
+        choices=contract["protocolVersions"],
+        default=contract["protocolVersions"],
+    )
+    parser.add_argument("--timeout", type=float, default=10.0)
+    parser.add_argument("--surface-only", action="store_true")
+    parser.add_argument("--output", type=Path)
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    contract = load_contract()
+    contract_issues = validate_contract(contract)
+    binary = args.binary.expanduser().resolve()
+    binary_issues: list[str] = []
+    if not binary.is_file():
+        binary_issues.append(f"binary does not exist: {binary}")
+    elif not os.access(binary, os.X_OK):
+        binary_issues.append(f"binary is not executable: {binary}")
+
+    probes: list[dict[str, Any]] = []
+    if not contract_issues and not binary_issues:
+        probes = [
+            probe_surface(binary, contract, pack, protocol, args.timeout)
+            for pack, protocol in matrix_jobs(contract, args.packs, args.protocols)
+        ]
+
+    fail_closed: dict[str, Any] | None = None
+    if not args.surface_only and not contract_issues and not binary_issues:
+        fail_closed = run_fail_closed_eval(
+            binary, contract["protocolVersions"][-1], args.timeout
+        )
+
+    passed = (
+        not contract_issues
+        and not binary_issues
+        and bool(probes)
+        and all(probe["status"] == "passed" for probe in probes)
+        and (fail_closed is None or fail_closed["passed"])
+    )
+    report = {
+        "schemaVersion": 1,
+        "status": "passed" if passed else "failed",
+        "binary": str(binary),
+        "contractIssues": contract_issues,
+        "binaryIssues": binary_issues,
+        "protocolMatrix": probes,
+        "failClosed": fail_closed,
+    }
+    rendered = json.dumps(report, indent=2, sort_keys=True) + "\n"
+    if args.output:
+        args.output.parent.mkdir(parents=True, exist_ok=True)
+        args.output.write_text(rendered, encoding="utf-8")
+    sys.stdout.write(rendered)
+    return 0 if passed else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/evals/tool-packs.json b/evals/tool-packs.json
new file mode 100644
index 0000000..5f9602f
--- /dev/null
+++ b/evals/tool-packs.json
@@ -0,0 +1,112 @@
+{
+  "schemaVersion": 1,
+  "defaultToolPack": "core",
+  "protocolVersions": [
+    "2024-11-05",
+    "2025-03-26",
+    "2025-06-18",
+    "2025-11-25",
+    "2026-07-28"
+  ],
+  "toolPacks": {
+    "core": [
+      "add_thought_node",
+      "consensus_answer",
+      "export_reasoning_audit",
+      "get_reasoning_coaching",
+      "reasoning_intent_router",
+      "record_assumption",
+      "run_reasoning_checkpoint",
+      "run_reasonkit_pipeline",
+      "run_thinking_mode",
+      "set_assumption_status",
+      "set_verification_policy",
+      "start_deliberation",
+      "verify_thoughts"
+    ],
+    "standard": [
+      "add_thought_node",
+      "apply_reasoning_pattern",
+      "consensus_answer",
+      "converge_reasoning",
+      "distill_thought_cluster",
+      "evaluate_reasoning_quality",
+      "expand_thoughts",
+      "export_memory_snapshot",
+      "export_reasoning_audit",
+      "get_reasoning_coaching",
+      "get_thinking_history",
+      "link_thoughts",
+      "list_reasoning_resources",
+      "merge_thought_branches",
+      "plan_tool_sequence",
+      "prune_thoughts",
+      "read_reasoning_resource",
+      "reasoning_intent_router",
+      "record_assumption",
+      "refine_thoughts",
+      "replay_reasoning_session",
+      "run_reasoning_checkpoint",
+      "run_reasonkit_pipeline",
+      "run_skeleton_elaboration",
+      "run_thinking_mode",
+      "score_thoughts",
+      "set_assumption_status",
+      "set_reasoning_aliases",
+      "set_scoring_rubric",
+      "set_verification_policy",
+      "start_deliberation",
+      "start_skeleton_of_thought",
+      "verify_thoughts"
+    ],
+    "full": [
+      "add_thought_node",
+      "apply_algorithm_template",
+      "apply_reasoning_pattern",
+      "clear_thinking_history",
+      "consensus_answer",
+      "converge_reasoning",
+      "distill_thought_cluster",
+      "evaluate_reasoning_quality",
+      "expand_thoughts",
+      "export_memory_snapshot",
+      "export_reasoning_audit",
+      "get_reasoning_coaching",
+      "get_thinking_history",
+      "link_thoughts",
+      "list_reasoning_resources",
+      "mcts_select_path",
+      "merge_thought_branches",
+      "plan_tool_sequence",
+      "prune_thoughts",
+      "query_failure_memory",
+      "read_reasoning_resource",
+      "reasoning_autopilot",
+      "reasoning_intent_router",
+      "record_assumption",
+      "record_reasoning_action",
+      "record_reasoning_observation",
+      "refine_thoughts",
+      "replay_reasoning_session",
+      "run_reasoning_checkpoint",
+      "run_reasonkit_pipeline",
+      "run_skeleton_elaboration",
+      "run_thinking_mode",
+      "score_thoughts",
+      "sequentialthinking_tools",
+      "set_assumption_status",
+      "set_reasoning_aliases",
+      "set_scoring_rubric",
+      "set_verification_policy",
+      "start_deliberation",
+      "start_skeleton_of_thought",
+      "think_query",
+      "verify_thoughts",
+      "web_research_enqueue",
+      "web_research_fetch",
+      "web_research_task_status",
+      "web_research_triangulate",
+      "web_research_verify"
+    ]
+  }
+}
diff --git a/examples/client-configs/claude-code.json b/examples/client-configs/claude-code.json
index 029df18..dee5c3d 100644
--- a/examples/client-configs/claude-code.json
+++ b/examples/client-configs/claude-code.json
@@ -1,10 +1,7 @@
 {
   "reasonkit-think": {
     "type": "stdio",
-    "command": "/absolute/path/to/reasonkit-think-mcp",
-    "args": [],
-    "env": {
-      "TMPDIR": "/tmp"
-    }
+    "command": "reasonkit-think-mcp",
+    "args": []
   }
 }
diff --git a/examples/client-configs/copilot-cli.json b/examples/client-configs/copilot-cli.json
index 710b54e..99abe28 100644
--- a/examples/client-configs/copilot-cli.json
+++ b/examples/client-configs/copilot-cli.json
@@ -2,11 +2,8 @@
   "mcpServers": {
     "reasonkit-think": {
       "type": "local",
-      "command": "/absolute/path/to/reasonkit-think-mcp",
+      "command": "reasonkit-think-mcp",
       "args": [],
-      "env": {
-        "TMPDIR": "/tmp"
-      },
       "tools": ["*"]
     }
   }
diff --git a/examples/client-configs/cursor.json b/examples/client-configs/cursor.json
index f09aaf9..affc72f 100644
--- a/examples/client-configs/cursor.json
+++ b/examples/client-configs/cursor.json
@@ -2,11 +2,8 @@
   "mcpServers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/absolute/path/to/reasonkit-think-mcp",
-      "args": [],
-      "env": {
-        "TMPDIR": "/tmp"
-      }
+      "command": "reasonkit-think-mcp",
+      "args": []
     }
   }
 }
diff --git a/examples/client-configs/generic-stdio.json b/examples/client-configs/generic-stdio.json
index 78c9d2f..affc72f 100644
--- a/examples/client-configs/generic-stdio.json
+++ b/examples/client-configs/generic-stdio.json
@@ -2,12 +2,8 @@
   "mcpServers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/absolute/path/to/reasonkit-think-mcp",
-      "args": [],
-      "env": {
-        "TMPDIR": "/tmp",
-        "RUST_LOG": "info"
-      }
+      "command": "reasonkit-think-mcp",
+      "args": []
     }
   }
 }
diff --git a/examples/client-configs/vscode-mcp.json b/examples/client-configs/vscode-mcp.json
index a4305ac..aa14994 100644
--- a/examples/client-configs/vscode-mcp.json
+++ b/examples/client-configs/vscode-mcp.json
@@ -2,11 +2,8 @@
   "servers": {
     "reasonkit-think": {
       "type": "stdio",
-      "command": "/absolute/path/to/reasonkit-think-mcp",
-      "args": [],
-      "env": {
-        "TMPDIR": "/tmp"
-      }
+      "command": "reasonkit-think-mcp",
+      "args": []
     }
   }
 }
diff --git a/features/README.md b/features/README.md
new file mode 100644
index 0000000..b7f6957
--- /dev/null
+++ b/features/README.md
@@ -0,0 +1,31 @@
+# Feature records
+
+Files under `done/` and `todo/` are planning records. Their directory is the
+record's editorial state, not an authoritative claim about the current runtime.
+Use this truth order when the sources disagree:
+
+1. Source behavior and executable tests.
+2. Exact discovery surfaces in `evals/tool-packs.json`.
+3. These feature records.
+
+[`manifest.json`](manifest.json) inventories every ID-bearing record. The
+community contract test rejects missing paths, duplicate IDs, state-directory
+mismatches, and headings that disagree with filenames.
+
+## Known overlap with the runtime
+
+| Record | Review | Current boundary |
+| --- | --- | --- |
+| T2-2A | Partial overlap | `apply_algorithm_template` creates deterministic scaffolds; it is not model-backed AoT. |
+| T2-2C | Partial overlap | `mcts_select_path` searches existing scored nodes without LLM rollouts. |
+| T2-2D | Partial overlap | `refine_thoughts` applies caller-supplied critique; it is not an autonomous Reflexion loop. |
+| T2-2E | Partial overlap | ReAct action and observation records exist, while the host still executes external tools. |
+| T2-2F | Proposal | No dedicated articulation or duck-panel tool is currently advertised. |
+| T3-3B | Partial overlap | Pipeline critique stages exist, but the proposed shadow-session protocol does not. |
+| T3-3C | Partial overlap | The Socratic reasoning pattern exists; dedicated question state and tools do not. |
+| T6-6B | Partial overlap | Failure patterns can be detected, persisted, and queried; the broader proposal remains a planning record. |
+| T7-P01–P10 | Partial overlap | `run_reasonkit_pipeline` supplies deterministic stage artifacts, not every acceptance criterion in the ten proposals. |
+
+Before moving a record, compare its acceptance criteria with tests and the
+pack-aware runtime contract. A matching tool name alone is not completion
+evidence.
diff --git a/features/manifest.json b/features/manifest.json
new file mode 100644
index 0000000..5b6aac7
--- /dev/null
+++ b/features/manifest.json
@@ -0,0 +1,150 @@
+{
+  "schemaVersion": 1,
+  "runtimeTruthOrder": [
+    "source and tests",
+    "evals/tool-packs.json",
+    "feature planning records"
+  ],
+  "records": [
+    {
+      "id": "T1-1A",
+      "path": "features/done/T1-1A-phase-1-complete.md",
+      "record_state": "done"
+    },
+    {
+      "id": "T1-1B",
+      "path": "features/done/T1-1B-semantic-scoring-rubric.md",
+      "record_state": "done"
+    },
+    {
+      "id": "T1-1C",
+      "path": "features/todo/T1-1C-embedding-diversity-pruning.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T2-2A",
+      "path": "features/todo/T2-2A-algorithm-of-thoughts-aot.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T2-2B",
+      "path": "features/done/T2-2B-skeleton-of-thoughts-sot.md",
+      "record_state": "done"
+    },
+    {
+      "id": "T2-2C",
+      "path": "features/todo/T2-2C-mcts-deliberation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T2-2D",
+      "path": "features/todo/T2-2D-reflexion-self-refine.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T2-2E",
+      "path": "features/todo/T2-2E-react-tool-augmented-reasoning.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T2-2F",
+      "path": "features/todo/T2-2F-rubber-duck-articulation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T3-3A",
+      "path": "features/done/T3-3A-meta-cognition-tool.md",
+      "record_state": "done"
+    },
+    {
+      "id": "T3-3B",
+      "path": "features/todo/T3-3B-adversarial-deliberation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T3-3C",
+      "path": "features/todo/T3-3C-socratic-questioning-engine.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T4-4A",
+      "path": "features/todo/T4-4A-bayesian-confidence-propagation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T4-4B",
+      "path": "features/todo/T4-4B-uncertainty-quantification-tags.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T5-5A",
+      "path": "features/done/T5-5A-new-edge-types.md",
+      "record_state": "done"
+    },
+    {
+      "id": "T5-5B",
+      "path": "features/todo/T5-5B-pagerank-importance-propagation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T6-6A",
+      "path": "features/todo/T6-6A-cross-session-pattern-extraction.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T6-6B",
+      "path": "features/todo/T6-6B-failure-mode-memory.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P01",
+      "path": "features/todo/T7-P01-pipeline-constraint-mapping.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P02",
+      "path": "features/todo/T7-P02-pipeline-diverse-framing.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P03",
+      "path": "features/todo/T7-P03-pipeline-hypothesis-generation.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P04",
+      "path": "features/todo/T7-P04-pipeline-evidence-gathering.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P05",
+      "path": "features/todo/T7-P05-pipeline-tradeoff-analysis.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P06",
+      "path": "features/todo/T7-P06-pipeline-adversarial-challenge.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P07",
+      "path": "features/todo/T7-P07-pipeline-synthesis.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P08",
+      "path": "features/todo/T7-P08-pipeline-verification.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P09",
+      "path": "features/todo/T7-P09-pipeline-metacognitive-review.md",
+      "record_state": "todo"
+    },
+    {
+      "id": "T7-P10",
+      "path": "features/todo/T7-P10-pipeline-confidence-routing.md",
+      "record_state": "todo"
+    }
+  ]
+}
diff --git a/features/todo/T2-2E-rubber-duck-articulation.md b/features/todo/T2-2F-rubber-duck-articulation.md
similarity index 99%
rename from features/todo/T2-2E-rubber-duck-articulation.md
rename to features/todo/T2-2F-rubber-duck-articulation.md
index 07acd78..ce9cc6b 100644
--- a/features/todo/T2-2E-rubber-duck-articulation.md
+++ b/features/todo/T2-2F-rubber-duck-articulation.md
@@ -1,4 +1,4 @@
-# T2-2E: Rubber Duck Articulation Tool
+# T2-2F: Rubber Duck Articulation Tool
 
 **Tier:** 2 — New Reasoning Paradigms  
 **Priority:** P1 (Highest impact on reasoning quality)  
@@ -381,4 +381,4 @@ PipelineStage::RubberDuckArticulation => {
 - Rubin, 1977 — "Mnemonic Vocabulary" — original rubber duck reference
 - Hunt & Thomas, 1999 — *The Pragmatic Programmer* — rubber duck debugging popularized
 - Madaan et al., 2023 — *Self-Refine* (arXiv:2303.17651) — iterative self-critique
-- nesquikm/mcp-rubber-duck — multi-LLM duck panel implementation
\ No newline at end of file
+- nesquikm/mcp-rubber-duck — multi-LLM duck panel implementation
diff --git a/justfile b/justfile
index 2087239..d7d95a1 100644
--- a/justfile
+++ b/justfile
@@ -4,21 +4,43 @@ default:
     @just --list
 
 build:
-    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{justfile_directory()}}/target}" cargo build --release
+    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}" cargo build --release
 
 check:
-    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{justfile_directory()}}/target}" cargo fmt --check
-    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{justfile_directory()}}/target}" cargo clippy -- -D warnings
-    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{justfile_directory()}}/target}" cargo test
+    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}" cargo fmt --check
+    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}" cargo clippy --all-targets --all-features -- -D warnings
+    just unit
+
+unit:
+    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}" cargo test --all-targets --all-features
+    python3 -m unittest discover -s tests -p "test_*.py" -v
+
+eval: build
+    python3 evals/run_contract_evals.py --binary "${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}/release/reasonkit-think-mcp"
+
+audit:
+    cargo audit --deny warnings
+
+package-check:
+    CARGO_TARGET_DIR="{{ justfile_directory() }}/target/package-check" cargo package --locked
+
+registry-check:
+    uvx --from check-jsonschema==0.38.0 check-jsonschema --schemafile https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json server.json
+
+dist-check:
+    @test "$(dist --version)" = "cargo-dist 0.32.0"
+    @dist manifest --artifacts=local --output-format=json --no-local-paths > /dev/null
+
+ci: check eval audit package-check registry-check dist-check
 
 fmt:
-    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{justfile_directory()}}/target}" cargo fmt
+    CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-{{ justfile_directory() }}/target}" cargo fmt
 
 # Rebuild release binary + validate workspace MCP wiring + smoke test.
 mcp-refresh:
     #!/usr/bin/env bash
     set -euo pipefail
-    ROOT="{{justfile_directory()}}"
+    ROOT="{{ justfile_directory() }}"
     CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-$ROOT/target}" cargo build --release
     python3 -m json.tool "$ROOT/../.mcp.json" >/dev/null
     test -x "$ROOT/target/release/reasonkit-think-mcp"
diff --git a/scripts/client_compat_check.py b/scripts/client_compat_check.py
index 3b7fef7..32cbb15 100755
--- a/scripts/client_compat_check.py
+++ b/scripts/client_compat_check.py
@@ -5,13 +5,22 @@
 
 import json
 import os
+import shutil
 import subprocess
 import sys
 import time
 from pathlib import Path
 
-PROTOCOL_VERSIONS = ("2024-11-05", "2025-03-26", "2025-06-18")
+PROTOCOL_VERSIONS = (
+    "2024-11-05",
+    "2025-03-26",
+    "2025-06-18",
+    "2025-11-25",
+    "2026-07-28",
+)
 EXPECTED_SERVER_NAME = "reasonkit-think-mcp"
+TOOL_PACK_COUNTS = {"core": 13, "standard": 33, "full": 47}
+VERSION_PROBE_TIMEOUT_SECONDS = 3
 
 
 def send(proc: subprocess.Popen[str], payload: dict) -> None:
@@ -56,10 +65,14 @@ def schema_type_issues(schema: object, path: str = "$") -> list[str]:
     return issues
 
 
-def probe_protocol(binary: Path, protocol_version: str) -> dict:
+def probe_protocol(binary: Path, protocol_version: str, tool_pack: str = "core") -> dict:
     env = os.environ.copy()
     env.setdefault("TMPDIR", "/tmp")
     env.setdefault("RUST_LOG", "warn")
+    if tool_pack == "core":
+        env.pop("REASONKIT_TOOL_PACK", None)
+    else:
+        env["REASONKIT_TOOL_PACK"] = tool_pack
     proc = subprocess.Popen(
         [str(binary)],
         cwd="/",
@@ -109,6 +122,7 @@ def probe_protocol(binary: Path, protocol_version: str) -> dict:
         server_info = initialize.get("result", {}).get("serverInfo", {})
         return {
             "protocol": protocol_version,
+            "tool_pack": tool_pack,
             "init_ms": init_ms,
             "server_name": server_info.get("name"),
             "tool_count": len(tools),
@@ -125,8 +139,108 @@ def probe_protocol(binary: Path, protocol_version: str) -> dict:
             proc.wait(timeout=2)
 
 
+def probe_invalid_tool_pack(binary: Path) -> dict:
+    env = os.environ.copy()
+    env.setdefault("TMPDIR", "/tmp")
+    env["RUST_LOG"] = "warn"
+    env["REASONKIT_TOOL_PACK"] = "invalid-pack"
+    proc = subprocess.Popen(
+        [str(binary)],
+        cwd="/",
+        env=env,
+        stdin=subprocess.PIPE,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        text=True,
+    )
+    tool_count = 0
+    stderr = ""
+    try:
+        send(
+            proc,
+            {
+                "jsonrpc": "2.0",
+                "id": 1,
+                "method": "initialize",
+                "params": {
+                    "protocolVersion": "2025-11-25",
+                    "capabilities": {},
+                    "clientInfo": {
+                        "name": "reasonkit-invalid-pack-check",
+                        "version": "1.0.0",
+                    },
+                },
+            },
+        )
+        recv(proc)
+        send(proc, {"jsonrpc": "2.0", "method": "notifications/initialized", "params": {}})
+        send(proc, {"jsonrpc": "2.0", "id": 2, "method": "tools/list", "params": {}})
+        tool_count = len(recv(proc).get("result", {}).get("tools", []))
+    finally:
+        proc.terminate()
+        try:
+            _, stderr = proc.communicate(timeout=2)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+            _, stderr = proc.communicate(timeout=2)
+    return {
+        "requested": "invalid-pack",
+        "fallback_tool_count": tool_count,
+        "warning_on_stderr": "invalid REASONKIT_TOOL_PACK" in stderr,
+    }
+
+
+def resolve_client_command(command: object) -> Path | None:
+    if not isinstance(command, str) or not command.strip():
+        return None
+
+    command = command.strip()
+    has_path_separator = os.sep in command or bool(os.altsep and os.altsep in command)
+    if has_path_separator:
+        return Path(command).expanduser().resolve()
+
+    resolved = shutil.which(command)
+    return Path(resolved).resolve() if resolved else None
+
+
+def executable_version(binary: Path | None) -> str | None:
+    if binary is None or not binary.is_file() or not os.access(binary, os.X_OK):
+        return None
+    try:
+        result = subprocess.run(
+            [str(binary), "--version"],
+            check=False,
+            capture_output=True,
+            text=True,
+            timeout=VERSION_PROBE_TIMEOUT_SECONDS,
+        )
+    except (OSError, subprocess.TimeoutExpired):
+        return None
+    if result.returncode != 0:
+        return None
+
+    first_line = result.stdout.strip().splitlines()
+    if not first_line:
+        return None
+    prefix = f"{EXPECTED_SERVER_NAME} "
+    if not first_line[0].startswith(prefix):
+        return None
+    return first_line[0].removeprefix(prefix).strip() or None
+
+
+def same_executable(left: Path | None, right: Path) -> bool:
+    if left is None or not left.exists() or not right.exists():
+        return False
+    try:
+        return left.samefile(right)
+    except OSError:
+        return left.resolve() == right.resolve()
+
+
 def client_config_checks(root: Path, binary: Path) -> list[dict]:
     home = Path.home()
+    candidate = binary.expanduser().resolve()
+    candidate_version = executable_version(candidate)
     configs = {
         "copilot-user": home / ".copilot" / "mcp-config.json",
         "copilot-workspace": root.parent / ".mcp.json",
@@ -149,14 +263,36 @@ def client_config_checks(root: Path, binary: Path) -> list[dict]:
             continue
         command = entry.get("command")
         type_ok = entry.get("type") in ("stdio", "local")
+        resolved = resolve_client_command(command)
+        command_exists = bool(resolved and resolved.is_file())
+        command_executable = bool(resolved and os.access(resolved, os.X_OK))
+        resolved_version = executable_version(resolved)
+        resolved_matches_candidate = same_executable(resolved, candidate)
+        version_matches_candidate = bool(
+            candidate_version
+            and resolved_version
+            and resolved_version == candidate_version
+        )
         checks.append(
             {
                 "client": name,
-                "status": "ok" if command == str(binary) and type_ok else "drift",
+                "status": (
+                    "ok"
+                    if type_ok
+                    and command_executable
+                    and (resolved_matches_candidate or version_matches_candidate)
+                    else "drift"
+                ),
                 "type": entry.get("type"),
+                "configured_command": command,
+                "resolved_command": str(resolved) if resolved else None,
                 "command_matches_release_binary": command == str(binary),
-                "command_exists": bool(command and Path(command).exists()),
-                "command_executable": bool(command and os.access(command, os.X_OK)),
+                "resolved_matches_release_binary": resolved_matches_candidate,
+                "candidate_version": candidate_version,
+                "resolved_version": resolved_version,
+                "version_matches_release_binary": version_matches_candidate,
+                "command_exists": command_exists,
+                "command_executable": command_executable,
             }
         )
     return checks
@@ -171,19 +307,25 @@ def main() -> int:
         "binary_exists": binary.exists(),
         "binary_executable": os.access(binary, os.X_OK),
         "protocols": [],
+        "invalid_tool_pack": {},
         "client_configs": client_config_checks(root, binary),
         "status": "failed",
     }
 
     if result["binary_exists"] and result["binary_executable"]:
-        result["protocols"] = [probe_protocol(binary, version) for version in PROTOCOL_VERSIONS]
+        result["protocols"] = [
+            probe_protocol(binary, version, tool_pack)
+            for version in PROTOCOL_VERSIONS
+            for tool_pack in TOOL_PACK_COUNTS
+        ]
+        result["invalid_tool_pack"] = probe_invalid_tool_pack(binary)
 
     ok = (
         result["binary_exists"]
         and result["binary_executable"]
         and all(
             protocol["server_name"] == EXPECTED_SERVER_NAME
-            and protocol["tool_count"] >= 30
+            and protocol["tool_count"] == TOOL_PACK_COUNTS[protocol["tool_pack"]]
             and protocol["prompt_count"] >= 8
             and protocol["schema_issue_count"] == 0
             for protocol in result["protocols"]
@@ -192,6 +334,8 @@ def main() -> int:
             check["status"] in ("ok", "missing_config")
             for check in result["client_configs"]
         )
+        and result["invalid_tool_pack"].get("fallback_tool_count") == 13
+        and result["invalid_tool_pack"].get("warning_on_stderr") is True
     )
     result["status"] = "ok" if ok else "failed"
     print(json.dumps(result, indent=2))
diff --git a/scripts/smoke_test.py b/scripts/smoke_test.py
index 8fcfbf6..7e54fb0 100755
--- a/scripts/smoke_test.py
+++ b/scripts/smoke_test.py
@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-"""End-to-end MCP stdio smoke test for reasonkit-think-mcp (v1 + v2)."""
+"""End-to-end MCP stdio smoke test for the complete full tool pack."""
 
 import json
 import os
@@ -43,11 +43,16 @@ def parse_resource_text(response: dict) -> dict:
 
 def main() -> int:
     root = Path(__file__).resolve().parent.parent
+    contract = json.loads((root / "evals" / "tool-packs.json").read_text())
+    expected_tools = contract["toolPacks"]["full"]
     for example in root.glob("examples/*.json"):
         json.loads(example.read_text())
 
     env = os.environ.copy()
-    env["CARGO_TARGET_DIR"] = str(root / "target")
+    env.setdefault("CARGO_TARGET_DIR", str(root / "target"))
+    # The comprehensive smoke exercises every compatibility route. The product
+    # default is the 13-tool core; full is the explicit 47-tool rollback pack.
+    env["REASONKIT_TOOL_PACK"] = "full"
 
     # Drain neither stderr here; a full `cargo run` build can fill PIPE and deadlock the child.
     proc = subprocess.Popen(
@@ -114,6 +119,24 @@ def main() -> int:
         )
         native_feature_triage = recv(proc)
 
+        send(
+            proc,
+            {
+                "jsonrpc": "2.0",
+                "id": 304,
+                "method": "resources/read",
+                "params": {"uri": "reasoning://config/tool-packs"},
+            },
+        )
+        native_tool_packs = recv(proc)
+
+        think_query_unavailable = call_tool(
+            proc,
+            305,
+            "think_query",
+            {"query": "must not simulate retrieval"},
+        )
+
         # v1 checks
         v1_ok = call_tool(
             proc,
@@ -294,7 +317,7 @@ def main() -> int:
             {"deliberation_id": harden_id},
         )
         harden_frontier = parse_text_result(harden_boot)["payload"]["frontier"]
-        call_tool(
+        harden_expand = call_tool(
             proc,
             21,
             "expand_thoughts",
@@ -304,6 +327,7 @@ def main() -> int:
                 "count": 1,
             },
         )
+        harden_scaffold_id = parse_text_result(harden_expand)["created_nodes"][0]["node_id"]
         harden_verify = call_tool(
             proc,
             22,
@@ -349,12 +373,44 @@ def main() -> int:
                 },
             },
         )
-        harden_consensus_relaxed = call_tool(
+        harden_action_spoof = call_tool(
+            proc,
+            250,
+            "record_reasoning_action",
+            {
+                "deliberation_id": harden_id,
+                "tool_name": "rg",
+                "parameters": {"pattern": "release"},
+                "tags": ["agent-authored"],
+                "add_to_frontier": True,
+            },
+        )
+        harden_scaffold_consensus = call_tool(
             proc,
             25,
             "consensus_answer",
             {"deliberation_id": harden_id},
         )
+        harden_agent_node = call_tool(
+            proc,
+            251,
+            "add_thought_node",
+            {
+                "deliberation_id": harden_id,
+                "content": "Agent-authored assessment accepts the residual claim risk under the explicitly relaxed policy.",
+                "parent_node_id": harden_scaffold_id,
+                "edge_kind": "refines",
+                "edge_note": "replace scaffold with host-authored assessment",
+                "tags": ["smoke", "governance-authorship"],
+                "add_to_frontier": True,
+            },
+        )
+        harden_consensus_relaxed = call_tool(
+            proc,
+            252,
+            "consensus_answer",
+            {"deliberation_id": harden_id},
+        )
         alias_set = call_tool(
             proc,
             26,
@@ -579,6 +635,16 @@ def main() -> int:
                 "assumption_id": assumption_payload["assumption"]["assumption_id"],
                 "status": "verified",
                 "confidence": 0.95,
+                "evidence": [
+                    {
+                        "source": "smoke-test-harness",
+                        "tier": "tier1",
+                        "independence_group": "smoke-test-harness",
+                        "supports": True,
+                        "contradictory": False,
+                        "unambiguous": True,
+                    }
+                ],
                 "notes": "smoke-test confirmation",
             },
         )
@@ -782,40 +848,6 @@ def main() -> int:
             },
         )
 
-        expected_tools = {
-            "sequentialthinking_tools",
-            "get_thinking_history",
-            "clear_thinking_history",
-            "start_deliberation",
-            "expand_thoughts",
-            "score_thoughts",
-            "prune_thoughts",
-            "start_skeleton_of_thought",
-            "run_skeleton_elaboration",
-            "link_thoughts",
-            "merge_thought_branches",
-            "distill_thought_cluster",
-            "refine_thoughts",
-            "record_assumption",
-            "set_assumption_status",
-            "replay_reasoning_session",
-            "converge_reasoning",
-            "apply_reasoning_pattern",
-            "verify_thoughts",
-            "consensus_answer",
-            "run_reasonkit_pipeline",
-            "run_reasoning_checkpoint",
-            "plan_tool_sequence",
-            "export_memory_snapshot",
-            "export_reasoning_audit",
-            "list_reasoning_resources",
-            "read_reasoning_resource",
-            "set_verification_policy",
-            "set_reasoning_aliases",
-            "reasoning_autopilot",
-            "reasoning_intent_router",
-            "run_thinking_mode",
-        }
         expected_prompts = {
             "sequential-thinking-guidance",
             "tot-planner-guidance",
@@ -839,6 +871,9 @@ def main() -> int:
         export_payload = parse_text_result(exported)
         graph_history_payload = parse_text_result(graph_history)
         harden_verify_payload = parse_text_result(harden_verify)
+        harden_action_spoof_payload = parse_text_result(harden_action_spoof)
+        harden_scaffold_consensus_payload = parse_text_result(harden_scaffold_consensus)
+        harden_consensus_relaxed_payload = parse_text_result(harden_consensus_relaxed)
         refined_got_payload = parse_text_result(refined_got)
         resource_list_payload = parse_text_result(resources_list)
         provider_payload = parse_text_result(providers_resource)
@@ -867,21 +902,34 @@ def main() -> int:
         sot_router_payload = parse_text_result(sot_router)
         native_thinking_modes_payload = parse_resource_text(native_thinking_modes)
         native_feature_triage_payload = parse_resource_text(native_feature_triage)
+        native_tool_packs_payload = parse_resource_text(native_tool_packs)
+        think_query_unavailable_payload = parse_text_result(think_query_unavailable)
         run_mode_payload = parse_text_result(run_mode)
         feature_router_payload = parse_text_result(feature_router)
         autopilot_map_payload = parse_text_result(autopilot_map)
 
         ok = (
             "result" in initialize
-            and expected_tools.issubset(set(tools))
+            and tools == expected_tools
             and expected_prompts.issubset(set(prompts))
             and "reasoning://thinking-modes" in native_resources
+            and "reasoning://config/tool-packs" in native_resources
             and "reasoning://workflows/feature-triage" in native_resources
             and native_thinking_modes_payload.get("default_tool") == "run_thinking_mode"
             and native_thinking_modes_payload.get("default_mode") == "auto"
+            and native_thinking_modes_payload.get("active_tool_pack") == "full"
+            and native_tool_packs_payload.get("active") == "full"
+            and native_tool_packs_payload.get("default") == "core"
+            and native_tool_packs_payload.get("active_tool_count") == 47
+            and think_query_unavailable.get("result", {}).get("isError", False)
+            and think_query_unavailable_payload.get("code")
+            == "experimental_memory_unavailable"
+            and think_query_unavailable.get("result", {}).get("structuredContent")
+            == think_query_unavailable_payload
             and "recommended_sequence" in native_feature_triage_payload
             and "does not read local files" in native_feature_triage_payload.get("critical_boundary", "")
             and not v1_ok.get("result", {}).get("isError", False)
+            and v1_ok.get("result", {}).get("structuredContent") == parse_text_result(v1_ok)
             and not v1_clear.get("result", {}).get("isError", False)
             and invalid.get("result", {}).get("isError", False)
             and invalid_payload.get("history_length") == 0
@@ -947,8 +995,17 @@ def main() -> int:
             and harden_verify_payload["claim_status_summary"]["data_deficit"] >= 1
             and harden_consensus.get("result", {}).get("isError", False)
             and not harden_policy_set.get("result", {}).get("isError", False)
+            and harden_action_spoof.get("result", {}).get("isError", False)
+            and harden_action_spoof_payload.get("code") == "invalid_input"
+            and harden_scaffold_consensus.get("result", {}).get("isError", False)
+            and harden_scaffold_consensus_payload.get("code") == "consensus_blocked"
+            and "final_answer" not in harden_scaffold_consensus_payload
+            and not harden_agent_node.get("result", {}).get("isError", False)
             and not alias_set.get("result", {}).get("isError", False)
             and not harden_consensus_relaxed.get("result", {}).get("isError", False)
+            and "Agent-authored assessment" in harden_consensus_relaxed_payload["final_answer"]
+            and "SCAFFOLD" not in harden_consensus_relaxed_payload["final_answer"]
+            and "hard fail check" not in harden_consensus_relaxed_payload["final_answer"]
             and "aliases" in parse_text_result(aliases_resource)
         )
 
@@ -958,6 +1015,8 @@ def main() -> int:
             "prompts": prompts,
             "native_resources": native_resources,
             "native_thinking_modes_default_tool": native_thinking_modes_payload.get("default_tool"),
+            "native_tool_pack": native_tool_packs_payload.get("active"),
+            "think_query_error": think_query_unavailable_payload.get("code"),
             "native_feature_triage_steps": len(native_feature_triage_payload.get("recommended_sequence", [])),
             "providers_active_mode": provider_payload.get("active_mode"),
             "telemetry_graph_depth_max": telemetry_payload.get("graph_depth_max"),
@@ -976,9 +1035,21 @@ def main() -> int:
             "harden_data_deficit": harden_verify_payload["claim_status_summary"]["data_deficit"],
             "harden_consensus_error": harden_consensus.get("result", {}).get("isError"),
             "harden_policy_set": not harden_policy_set.get("result", {}).get("isError", False),
+            "harden_action_spoof_error": harden_action_spoof.get("result", {}).get(
+                "isError"
+            ),
+            "harden_scaffold_consensus_error": harden_scaffold_consensus.get(
+                "result", {}
+            ).get("isError"),
+            "harden_agent_node": not harden_agent_node.get("result", {}).get("isError", False),
             "aliases_set": not alias_set.get("result", {}).get("isError", False),
             "harden_relaxed_consensus_error": harden_consensus_relaxed.get("result", {}).get("isError"),
-            "harden_relaxed_consensus_payload": harden_consensus_relaxed.get("result", {}),
+            "harden_relaxed_consensus_route": harden_consensus_relaxed_payload.get(
+                "route_decision"
+            ),
+            "harden_relaxed_consensus_answer": harden_consensus_relaxed_payload.get(
+                "final_answer"
+            ),
             "sot_phase_id": sot_phase_id,
             "sot_elaboration_status": sot_elaboration_payload.get("status"),
             "sot_router_tools": sot_router_payload.get("tool_plan"),
diff --git a/server.json b/server.json
new file mode 100644
index 0000000..d10b6c5
--- /dev/null
+++ b/server.json
@@ -0,0 +1,21 @@
+{
+  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
+  "name": "io.github.reasonkit/reasonkit-think",
+  "title": "ReasonKit Think",
+  "description": "Auditable decision protocol and fail-closed evidence gate for MCP agents",
+  "version": "0.2.0",
+  "repository": {
+    "url": "https://github.com/reasonkit/ReasonKit-think",
+    "source": "github"
+  },
+  "packages": [
+    {
+      "registryType": "cargo",
+      "identifier": "reasonkit-think-mcp",
+      "version": "0.2.0",
+      "transport": {
+        "type": "stdio"
+      }
+    }
+  ]
+}
diff --git a/src/agent_protocol.rs b/src/agent_protocol.rs
index 551374d..d6b9c11 100644
--- a/src/agent_protocol.rs
+++ b/src/agent_protocol.rs
@@ -20,6 +20,7 @@ pub struct NodeView {
     pub score: Option<f64>,
     pub branch_id: Option<String>,
     pub node_type: ThoughtNodeType,
+    pub agent_authored_provenance: bool,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize, Default, JsonSchema)]
@@ -134,7 +135,9 @@ const SCAFFOLD_TAGS: &[&str] = &[
 ];
 
 pub fn is_agent_authored(node: &NodeView) -> bool {
-    node.tags.iter().any(|t| t == "agent-authored")
+    matches!(node.node_type, ThoughtNodeType::Thought)
+        && node.agent_authored_provenance
+        && node.tags.iter().any(|t| t == "agent-authored")
 }
 
 pub fn is_scaffold_node(node: &NodeView) -> bool {
@@ -356,8 +359,11 @@ pub struct PipelineContext<'a> {
     pub blocking_assumptions: &'a [String],
 }
 
-pub fn run_pipeline_stage(stage: &str, ctx: &PipelineContext<'_>) -> PipelineStageArtifact {
-    match stage {
+pub fn run_pipeline_stage(
+    stage: &str,
+    ctx: &PipelineContext<'_>,
+) -> Result<PipelineStageArtifact, String> {
+    let artifact = match stage {
         "constraint-mapping" => stage_constraint_mapping(ctx),
         "diverse-thinking" => stage_diverse_thinking(ctx),
         "first-principles" => stage_first_principles(ctx),
@@ -368,8 +374,9 @@ pub fn run_pipeline_stage(stage: &str, ctx: &PipelineContext<'_>) -> PipelineSta
         "triangulation" => stage_triangulation(ctx),
         "brutal-honesty" => stage_brutal_honesty(ctx),
         "calibration-routing" => stage_calibration_routing(ctx),
-        _ => stage_generic(stage, ctx),
-    }
+        _ => return Err(format!("invalid pipeline stage id: {stage}")),
+    };
+    Ok(artifact)
 }
 
 fn stage_constraint_mapping(ctx: &PipelineContext<'_>) -> PipelineStageArtifact {
@@ -684,23 +691,6 @@ fn stage_calibration_routing(ctx: &PipelineContext<'_>) -> PipelineStageArtifact
     }
 }
 
-fn stage_generic(stage: &str, ctx: &PipelineContext<'_>) -> PipelineStageArtifact {
-    PipelineStageArtifact {
-        stage: stage.to_string(),
-        finding: format!("Stage `{stage}` completed (generic protocol marker)"),
-        confidence: 0.6,
-        confidence_basis: "Unknown stage id; minimal structural pass".to_string(),
-        structured: json!({ "goal": ctx.goal }),
-        gate: StageGate {
-            stage: stage.to_string(),
-            passed: true,
-            blockers: Vec::new(),
-            recommended_tools: vec!["add_thought_node".to_string()],
-        },
-        agent_tasks: Vec::new(),
-    }
-}
-
 /// Structural MCTS over existing scored nodes (no LLM rollouts).
 pub fn mcts_recommend_path(
     nodes: &HashMap<String, NodeView>,
@@ -907,6 +897,7 @@ mod tests {
             score: None,
             branch_id: Some("b1".into()),
             node_type: ThoughtNodeType::Thought,
+            agent_authored_provenance: false,
         };
         assert!(is_scaffold_node(&n));
         assert!(!is_agent_authored(&n));
diff --git a/src/main.rs b/src/main.rs
index f911396..eff3faf 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -7,7 +7,10 @@ use std::sync::atomic::{AtomicU64, Ordering};
 use std::sync::{Arc, Mutex};
 use std::time::Duration;
 
-use rmcp::handler::server::{router::prompt::PromptRouter, wrapper::Parameters};
+use rmcp::handler::server::{
+    router::{prompt::PromptRouter, tool::ToolRouter},
+    wrapper::Parameters,
+};
 use rmcp::{
     ErrorData as McpError, RoleServer,
     model::*,
@@ -39,6 +42,174 @@ const REDACTION: &str = "[redacted: possible prompt-injection text]";
 
 static ID_SEQ: AtomicU64 = AtomicU64::new(1);
 
+const CORE_TOOL_NAMES: &[&str] = &[
+    "add_thought_node",
+    "consensus_answer",
+    "export_reasoning_audit",
+    "get_reasoning_coaching",
+    "reasoning_intent_router",
+    "record_assumption",
+    "run_reasoning_checkpoint",
+    "run_reasonkit_pipeline",
+    "run_thinking_mode",
+    "set_assumption_status",
+    "set_verification_policy",
+    "start_deliberation",
+    "verify_thoughts",
+];
+
+const STANDARD_TOOL_NAMES: &[&str] = &[
+    "add_thought_node",
+    "apply_reasoning_pattern",
+    "consensus_answer",
+    "converge_reasoning",
+    "distill_thought_cluster",
+    "evaluate_reasoning_quality",
+    "expand_thoughts",
+    "export_memory_snapshot",
+    "export_reasoning_audit",
+    "get_reasoning_coaching",
+    "get_thinking_history",
+    "link_thoughts",
+    "list_reasoning_resources",
+    "merge_thought_branches",
+    "plan_tool_sequence",
+    "prune_thoughts",
+    "read_reasoning_resource",
+    "reasoning_intent_router",
+    "record_assumption",
+    "refine_thoughts",
+    "replay_reasoning_session",
+    "run_reasoning_checkpoint",
+    "run_reasonkit_pipeline",
+    "run_skeleton_elaboration",
+    "run_thinking_mode",
+    "score_thoughts",
+    "set_assumption_status",
+    "set_reasoning_aliases",
+    "set_scoring_rubric",
+    "set_verification_policy",
+    "start_deliberation",
+    "start_skeleton_of_thought",
+    "verify_thoughts",
+];
+
+const FULL_TOOL_NAMES: &[&str] = &[
+    "add_thought_node",
+    "apply_algorithm_template",
+    "apply_reasoning_pattern",
+    "clear_thinking_history",
+    "consensus_answer",
+    "converge_reasoning",
+    "distill_thought_cluster",
+    "evaluate_reasoning_quality",
+    "expand_thoughts",
+    "export_memory_snapshot",
+    "export_reasoning_audit",
+    "get_reasoning_coaching",
+    "get_thinking_history",
+    "link_thoughts",
+    "list_reasoning_resources",
+    "mcts_select_path",
+    "merge_thought_branches",
+    "plan_tool_sequence",
+    "prune_thoughts",
+    "query_failure_memory",
+    "read_reasoning_resource",
+    "reasoning_autopilot",
+    "reasoning_intent_router",
+    "record_assumption",
+    "record_reasoning_action",
+    "record_reasoning_observation",
+    "refine_thoughts",
+    "replay_reasoning_session",
+    "run_reasoning_checkpoint",
+    "run_reasonkit_pipeline",
+    "run_skeleton_elaboration",
+    "run_thinking_mode",
+    "score_thoughts",
+    "sequentialthinking_tools",
+    "set_assumption_status",
+    "set_reasoning_aliases",
+    "set_scoring_rubric",
+    "set_verification_policy",
+    "start_deliberation",
+    "start_skeleton_of_thought",
+    "think_query",
+    "verify_thoughts",
+    "web_research_enqueue",
+    "web_research_fetch",
+    "web_research_task_status",
+    "web_research_triangulate",
+    "web_research_verify",
+];
+
+const INVALID_TOOL_PACK_WARNING: &str =
+    "invalid REASONKIT_TOOL_PACK; expected core, standard, or full; using safe core fallback";
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize)]
+#[serde(rename_all = "snake_case")]
+enum ToolPack {
+    Core,
+    Standard,
+    Full,
+}
+
+impl ToolPack {
+    fn as_str(self) -> &'static str {
+        match self {
+            Self::Core => "core",
+            Self::Standard => "standard",
+            Self::Full => "full",
+        }
+    }
+
+    fn members(self) -> &'static [&'static str] {
+        match self {
+            Self::Core => CORE_TOOL_NAMES,
+            Self::Standard => STANDARD_TOOL_NAMES,
+            Self::Full => FULL_TOOL_NAMES,
+        }
+    }
+
+    fn contains(self, tool_name: &str) -> bool {
+        self.members().binary_search(&tool_name).is_ok()
+    }
+
+    fn minimum_for(tool_name: &str) -> Option<Self> {
+        if Self::Core.contains(tool_name) {
+            Some(Self::Core)
+        } else if Self::Standard.contains(tool_name) {
+            Some(Self::Standard)
+        } else if Self::Full.contains(tool_name) {
+            Some(Self::Full)
+        } else {
+            None
+        }
+    }
+
+    fn resolve(raw: Option<&str>) -> (Self, Option<&'static str>) {
+        match raw.map(str::trim) {
+            None | Some("") | Some("core") => (Self::Core, None),
+            Some("standard") => (Self::Standard, None),
+            Some("full") => (Self::Full, None),
+            Some(_) => (Self::Core, Some(INVALID_TOOL_PACK_WARNING)),
+        }
+    }
+
+    fn from_env() -> Self {
+        let (pack, warning) = match std::env::var("REASONKIT_TOOL_PACK") {
+            Ok(value) => Self::resolve(Some(&value)),
+            Err(std::env::VarError::NotPresent) => Self::resolve(None),
+            Err(std::env::VarError::NotUnicode(_)) => (Self::Core, Some(INVALID_TOOL_PACK_WARNING)),
+        };
+        if let Some(message) = warning {
+            tracing::warn!("{message}");
+        }
+        pack
+    }
+}
+
 macro_rules! lock_or_return {
     ($lock:expr, $store_name:literal) => {
         match $lock.lock() {
@@ -48,27 +219,114 @@ macro_rules! lock_or_return {
     };
 }
 
-/// Backend mode for reasoning and evaluation.
-/// - Agent (default): AI agent does all reasoning locally using built-in capabilities
-/// - External (future): AI agent can delegate to external LLM APIs for scoring/peer-review
+const INVALID_BACKEND_MODE_WARNING: &str =
+    "invalid BACKEND_MODE; expected agent or external; using safe agent fallback";
+const EXTERNAL_BACKEND_UNAVAILABLE_WARNING: &str = "BACKEND_MODE=external is unavailable because no external provider is bound; using safe agent fallback";
+
+/// Effective backend mode for reasoning and evaluation.
+/// The host agent is the only bound backend; reserved external configuration falls back here.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default)]
 enum BackendMode {
     #[serde(rename = "agent")]
     #[default]
     Agent,
-    #[serde(rename = "external")]
-    External,
 }
 
 impl BackendMode {
+    fn resolve(raw: Option<&str>) -> (Self, Option<&'static str>) {
+        match raw.map(str::trim) {
+            None | Some("agent") => (Self::Agent, None),
+            Some("external") => (Self::Agent, Some(EXTERNAL_BACKEND_UNAVAILABLE_WARNING)),
+            Some(_) => (Self::Agent, Some(INVALID_BACKEND_MODE_WARNING)),
+        }
+    }
+
+    fn resolve_env() -> (Self, Option<&'static str>) {
+        match std::env::var("BACKEND_MODE") {
+            Ok(value) => Self::resolve(Some(&value)),
+            Err(std::env::VarError::NotPresent) => Self::resolve(None),
+            Err(std::env::VarError::NotUnicode(_)) => {
+                (Self::Agent, Some(INVALID_BACKEND_MODE_WARNING))
+            }
+        }
+    }
+
     fn from_env() -> Self {
-        match std::env::var("BACKEND_MODE").as_deref() {
-            Ok("external") => BackendMode::External,
-            _ => BackendMode::Agent,
+        let (mode, warning) = Self::resolve_env();
+        if let Some(message) = warning {
+            tracing::warn!("{message}");
+        }
+        mode
+    }
+
+    fn as_str(self) -> &'static str {
+        match self {
+            Self::Agent => "agent",
         }
     }
 }
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum CliAction {
+    ServeStdio,
+    Version,
+    Help,
+    DoctorJson,
+}
+
+const CLI_HELP: &str = "ReasonKit Think MCP — governed reasoning protocols for AI agents
+
+Usage:
+  reasonkit-think-mcp                 Start the MCP stdio server
+  reasonkit-think-mcp doctor --json   Print a machine-readable installation report
+  reasonkit-think-mcp --version       Print the installed version
+  reasonkit-think-mcp --help          Print this help
+
+Environment:
+  REASONKIT_TOOL_PACK=core|standard|full  Advertised tool surface (default: core)
+  BACKEND_MODE=agent|external             Reasoning backend; external currently warns and falls back to agent";
+
+fn parse_cli_action(args: &[String]) -> Result<CliAction, String> {
+    match args {
+        [] => Ok(CliAction::ServeStdio),
+        [arg] if arg == "--version" || arg == "-V" => Ok(CliAction::Version),
+        [arg] if arg == "--help" || arg == "-h" => Ok(CliAction::Help),
+        [command] if command == "doctor" => Ok(CliAction::DoctorJson),
+        [command, format] if command == "doctor" && format == "--json" => Ok(CliAction::DoctorJson),
+        _ => Err(format!(
+            "unsupported arguments: {}\n\n{CLI_HELP}",
+            args.join(" ")
+        )),
+    }
+}
+
+fn doctor_report(
+    raw_tool_pack: Option<&str>,
+    backend_resolution: (BackendMode, Option<&'static str>),
+) -> Value {
+    let (tool_pack, tool_pack_warning) = ToolPack::resolve(raw_tool_pack);
+    let (backend_mode, backend_warning) = backend_resolution;
+    let warnings = [tool_pack_warning, backend_warning]
+        .into_iter()
+        .flatten()
+        .collect::<Vec<_>>();
+
+    json!({
+        "schema_version": 1,
+        "status": if warnings.is_empty() { "ok" } else { "warning" },
+        "version": env!("CARGO_PKG_VERSION"),
+        "transport": "stdio",
+        "active_tool_pack": tool_pack.as_str(),
+        "tool_count": tool_pack.members().len(),
+        "semantic_engine": "host_agent",
+        "backend_mode": backend_mode.as_str(),
+        "external_provider_bound": false,
+        "retrieval_provider_bound": false,
+        "structured_content": true,
+        "warnings": warnings,
+    })
+}
+
 #[derive(Debug, Serialize, Deserialize)]
 struct AppState {
     thoughts: ThinkingStore,
@@ -296,9 +554,10 @@ struct ThinkServer {
     deliberations: Arc<Mutex<DeliberationStore>>,
     failure_memory: Arc<Mutex<HashMap<String, FailureModeRecord>>>,
     session_pattern_notes: Arc<Mutex<Vec<String>>>,
-    hot_memory: Arc<reasonkit_mem::storage::hot::HotMemory>,
     #[allow(dead_code)]
     backend_mode: BackendMode,
+    tool_router: ToolRouter<Self>,
+    tool_pack: ToolPack,
     #[allow(dead_code)]
     prompt_router: PromptRouter<Self>,
     obs: Observability,
@@ -511,6 +770,10 @@ impl ThinkServer {
     }
 
     fn new(max_history_size: usize) -> Self {
+        Self::new_with_tool_pack(max_history_size, ToolPack::from_env())
+    }
+
+    fn new_with_tool_pack(max_history_size: usize, tool_pack: ToolPack) -> Self {
         let path = Self::state_path().ok();
         let mut thoughts_init = ThinkingStore::new(max_history_size);
         let mut deliberations_init = DeliberationStore::new(max_history_size);
@@ -538,22 +801,61 @@ impl ThinkServer {
             "ReasonKit-think protocol mode (no server-side LLM reasoning)"
         );
 
-        let hot_memory = Arc::new(reasonkit_mem::storage::hot::HotMemory::new(
-            reasonkit_mem::storage::hot::HotMemoryConfig::default(),
-        ));
+        let mut tool_router = Self::tool_router();
+        for tool in tool_router.list_all() {
+            if !tool_pack.contains(tool.name.as_ref()) {
+                let disabled = tool_router.disable_route(tool.name);
+                debug_assert!(disabled, "generated tool route must exist");
+            }
+        }
+        assert_eq!(
+            tool_router.list_all().len(),
+            tool_pack.members().len(),
+            "tool-pack contract drift: configured names must match generated routes"
+        );
+        tracing::info!(
+            tool_pack = tool_pack.as_str(),
+            tool_count = tool_pack.members().len(),
+            "ReasonKit-think tool pack selected"
+        );
 
         Self {
             thoughts: Arc::new(Mutex::new(thoughts_init)),
             deliberations: Arc::new(Mutex::new(deliberations_init)),
             failure_memory: Arc::new(Mutex::new(failure_memory_init)),
             session_pattern_notes: Arc::new(Mutex::new(session_pattern_notes_init)),
-            hot_memory,
             backend_mode,
+            tool_router,
+            tool_pack,
             prompt_router: Self::prompt_router(),
             obs: Observability::new(),
         }
     }
 
+    fn with_tool_pack_guidance(&self, text: String, tools: &[&str]) -> String {
+        let unavailable = tools
+            .iter()
+            .filter(|tool| !self.tool_pack.contains(tool))
+            .filter_map(|tool| {
+                ToolPack::minimum_for(tool)
+                    .map(|pack| format!("{tool} requires the {} tool pack", pack.as_str()))
+            })
+            .collect::<Vec<_>>();
+        let guidance = if unavailable.is_empty() {
+            format!(
+                "Active tool pack: {}. All tools named by this prompt are available.",
+                self.tool_pack.as_str()
+            )
+        } else {
+            format!(
+                "Active tool pack: {}. {}. Use run_thinking_mode for the core workflow or restart with the required REASONKIT_TOOL_PACK value.",
+                self.tool_pack.as_str(),
+                unavailable.join("; ")
+            )
+        };
+        format!("{text}\n\nTool-pack guidance: {guidance}")
+    }
+
     fn run_thinking_mode_inner(
         &self,
         input: RunThinkingModeInput,
@@ -814,6 +1116,7 @@ impl ThinkServer {
 
         let audit = store.export_audit(ExportReasoningAuditInput {
             deliberation_id: deliberation_id.clone(),
+            format: AuditFormat::Json,
             include_raw_thoughts: Some(false),
         })?;
         executed_tools.push("export_reasoning_audit".to_string());
@@ -1181,6 +1484,7 @@ impl ThinkServer {
 
                 let audit = store.export_audit(ExportReasoningAuditInput {
                     deliberation_id: deliberation_id.clone(),
+                    format: AuditFormat::Json,
                     include_raw_thoughts: Some(true),
                 })?;
                 executed.push("export_reasoning_audit".to_string());
@@ -1920,7 +2224,7 @@ exactly what weights are active after normalization."
             Ok(out) => {
                 let persist_error = self.save_state().err();
                 if out.policy_blocked {
-                    json_tool_error(&out)
+                    consensus_blocked_tool_result(&out)
                 } else if let Some(err) = persist_error {
                     json_tool_error(&ErrorEnvelope::new("persistence_error", err, true))
                 } else {
@@ -2011,7 +2315,9 @@ exactly what weights are active after normalization."
         }
     }
 
-    #[tool(description = "Export a complete reasoning audit artifact for a deliberation.")]
+    #[tool(
+        description = "Export the canonical reasoning audit payload plus a deterministic JSON or Markdown artifact. Raw thought content is omitted unless explicitly requested."
+    )]
     fn export_reasoning_audit(
         &self,
         Parameters(input): Parameters<ExportReasoningAuditInput>,
@@ -2276,6 +2582,9 @@ exactly what weights are active after normalization."
         if input.uri == "reasoning://config/providers" {
             return json_tool_success(&provider_capabilities());
         }
+        if input.uri == "reasoning://config/tool-packs" {
+            return json_tool_success(&tool_pack_resource(self.tool_pack));
+        }
         if input.uri == "reasoning://config/web-research" {
             return json_tool_success(&json!({
                 "base_url": Self::web_research_base_url(),
@@ -2303,79 +2612,24 @@ exactly what weights are active after normalization."
             return json_tool_success(&memory_sinks_resource());
         }
         let store = lock_or_return!(self.deliberations, "deliberation");
-        match read_reasoning_resource_payload(&store, &input.uri) {
+        match read_reasoning_resource_payload(&store, &input.uri, self.tool_pack) {
             Ok(payload) => json_tool_success(&payload),
             Err(err) => json_tool_error(&ErrorEnvelope::from_store_error(err)),
         }
     }
 
     #[tool(
-        description = "Perform a blazing fast semantic query using the hot-tier memory cache, integrating BM25 + vector search and RAPTOR expansion without writing to persistent stores."
+        description = "Compatibility-only experimental memory query. No production provider is bound, so this tool fails closed with experimental_memory_unavailable and performs no retrieval."
     )]
-    fn think_query(&self, Parameters(input): Parameters<ThinkQueryInput>) -> CallToolResult {
+    fn think_query(&self, Parameters(_input): Parameters<ThinkQueryInput>) -> CallToolResult {
         let started = std::time::Instant::now();
-        let cache = &self.hot_memory;
-
-        let embedding = input.query_embedding.unwrap_or_else(|| vec![0.0; 1536]);
-        let limit = input.limit.unwrap_or(10);
-        let min_score = input.min_score.unwrap_or(0.0);
-
-        let dense_results = tokio::task::block_in_place(|| {
-            tokio::runtime::Handle::current().block_on(async {
-                cache
-                    .search_similar_with_threshold(&embedding, limit, min_score)
-                    .await
-            })
-        });
-
-        // Initialize fusion engine
-        use reasonkit_mem::retrieval::fusion::{FusionEngine, to_ranked_results};
-        let fusion_engine = FusionEngine::rrf(60);
-        let mut results_map = std::collections::HashMap::new();
-
-        // Add dense results
-        results_map.insert("dense".to_string(), to_ranked_results(dense_results));
-
-        // TODO: In the future, add sparse (BM25) results here:
-        // results_map.insert("sparse".to_string(), to_ranked_results(sparse_results));
-
-        let fused_results = fusion_engine.fuse(results_map).unwrap_or_default();
-
-        let mut final_results = Vec::new();
-        for fused in fused_results.into_iter().take(limit) {
-            if let Some(entry) = cache.peek(&fused.id) {
-                final_results.push(HotMemoryEntryResult {
-                    id: entry.id.to_string(),
-                    content: entry.content,
-                    score: fused.fusion_score,
-                    metadata: entry.metadata,
-                });
-            }
-        }
-
-        let failure_coaching = self
-            .failure_memory
-            .lock()
-            .ok()
-            .map(|mem| {
-                let mut patterns: Vec<_> = mem.values().collect();
-                patterns.sort_by_key(|b| std::cmp::Reverse(b.occurrence_count));
-                patterns
-                    .into_iter()
-                    .take(5)
-                    .map(|p| format!("[{}] {}", p.pattern_id, p.intervention))
-                    .collect::<Vec<_>>()
-            })
-            .unwrap_or_default();
-
         self.obs
             .record_tool_latency("think_query", started.elapsed());
-        json_tool_success(&ThinkQueryResult {
-            results: final_results,
-            latency_ms: started.elapsed().as_millis(),
-            failure_coaching,
-            semantic_engine: "host_agent".to_string(),
-        })
+        json_tool_error(&ErrorEnvelope::new(
+            "experimental_memory_unavailable",
+            "think_query has no bound production memory provider; no retrieval was performed",
+            false,
+        ))
     }
 }
 
@@ -2398,10 +2652,8 @@ impl ThinkServer {
         if let Some(problem) = input.problem {
             lines.push(format!("Problem: {problem}"));
         }
-        GetPromptResult::new(vec![PromptMessage::new_text(
-            PromptMessageRole::User,
-            lines.join("\n"),
-        )])
+        let text = self.with_tool_pack_guidance(lines.join("\n"), &["sequentialthinking_tools"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2418,7 +2670,16 @@ impl ThinkServer {
         let text = format!(
             "Plan with explicit branches.\n1) Create 3-5 candidate branches.\n2) Score each on correctness, risk, and evidence.\n3) Prune weak branches but preserve diversity.\n4) Verify critical claims before final answer.\nProblem: {problem}"
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(
+            text,
+            &[
+                "expand_thoughts",
+                "score_thoughts",
+                "prune_thoughts",
+                "verify_thoughts",
+            ],
+        );
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2441,7 +2702,20 @@ impl ThinkServer {
 5) Score, verify critical claims, run a checkpoint, then converge with GoT merge/distill before consensus.\n\
 Problem: {problem}"
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(
+            text,
+            &[
+                "start_skeleton_of_thought",
+                "run_skeleton_elaboration",
+                "add_thought_node",
+                "score_thoughts",
+                "verify_thoughts",
+                "run_reasoning_checkpoint",
+                "converge_reasoning",
+                "consensus_answer",
+            ],
+        );
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2456,7 +2730,8 @@ Problem: {problem}"
             "Apply these lenses independently then synthesize: Optimist, Pessimist, Systems Thinker, Empiricist, Contrarian, Security Adversary, Simplifier.\nTopic: {}",
             input.problem.unwrap_or_else(|| "General topic".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["add_thought_node"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2471,7 +2746,8 @@ Problem: {problem}"
             "Use Chain-of-Verification: draft -> verification questions -> independent answers -> revised final.\nTarget: {}",
             input.problem.unwrap_or_else(|| "Current draft".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["verify_thoughts"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2488,7 +2764,8 @@ Problem: {problem}"
                 .problem
                 .unwrap_or_else(|| "Current answer".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["verify_thoughts"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2505,7 +2782,9 @@ Problem: {problem}"
                 .problem
                 .unwrap_or_else(|| "Current deliberation".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text =
+            self.with_tool_pack_guidance(text, &["run_reasoning_checkpoint", "consensus_answer"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2522,7 +2801,8 @@ Problem: {problem}"
                 .problem
                 .unwrap_or_else(|| "Current deliberation".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["expand_thoughts", "add_thought_node"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2539,7 +2819,11 @@ Problem: {problem}"
                 .problem
                 .unwrap_or_else(|| "Current deliberation".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(
+            text,
+            &["link_thoughts", "record_assumption", "add_thought_node"],
+        );
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2556,7 +2840,8 @@ Problem: {problem}"
                 .problem
                 .unwrap_or_else(|| "Current deliberation".to_string())
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["score_thoughts", "record_assumption"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 
     #[prompt(
@@ -2580,11 +2865,12 @@ Problem: {problem}"
 6) Final answer must name the winning feature, runner-up, rejected alternatives, exact evidence used, uncertainty, and route decision.\n\
 Do not fabricate feature names, TODO status, effort, dependencies, or implementation claims."
         );
-        GetPromptResult::new(vec![PromptMessage::new_text(PromptMessageRole::User, text)])
+        let text = self.with_tool_pack_guidance(text, &["run_thinking_mode"]);
+        GetPromptResult::new(vec![PromptMessage::new_text(Role::User, text)])
     }
 }
 
-#[tool_handler]
+#[tool_handler(router = self.tool_router)]
 #[prompt_handler]
 impl ServerHandler for ThinkServer {
     fn get_info(&self) -> ServerInfo {
@@ -2601,9 +2887,11 @@ impl ServerHandler for ThinkServer {
                 .build(),
         )
         .with_server_info(server_info)
-        .with_instructions(
-            "Use ReasonKit Think for auditable CoT, ToT, GoT, verification, governance, and reasoning-resource workflows over stdio.",
-        )
+        .with_instructions(format!(
+            "Use ReasonKit Think for auditable reasoning and fail-closed governance over stdio; active tool pack: {} ({} tools). Read reasoning://config/tool-packs for the exact surface. Restart with REASONKIT_TOOL_PACK=full for the 47-tool compatibility surface.",
+            self.tool_pack.as_str(),
+            self.tool_pack.members().len()
+        ))
     }
 
     fn list_resources(
@@ -2633,13 +2921,15 @@ impl ServerHandler for ThinkServer {
         &self,
         request: ReadResourceRequestParams,
         _context: RequestContext<RoleServer>,
-    ) -> impl std::future::Future<Output = Result<ReadResourceResult, McpError>> + MaybeSendFuture + '_
+    ) -> impl std::future::Future<Output = Result<ReadResourceResponse, McpError>> + MaybeSendFuture + '_
     {
         let uri = request.uri;
         let payload = if uri == "reasoning://telemetry/summary" {
             Ok(self.obs.snapshot())
         } else if uri == "reasoning://config/providers" {
             Ok(provider_capabilities())
+        } else if uri == "reasoning://config/tool-packs" {
+            Ok(tool_pack_resource(self.tool_pack))
         } else if uri == "reasoning://config/web-research" {
             Ok(json!({
                 "base_url": Self::web_research_base_url(),
@@ -2668,7 +2958,7 @@ impl ServerHandler for ThinkServer {
             self.deliberations
                 .lock()
                 .map_err(|_| "deliberation store lock unavailable".to_string())
-                .and_then(|store| read_reasoning_resource_payload(&store, &uri))
+                .and_then(|store| read_reasoning_resource_payload(&store, &uri, self.tool_pack))
         };
         let result = payload.and_then(|payload| {
             serde_json::to_string_pretty(&payload)
@@ -2677,7 +2967,8 @@ impl ServerHandler for ThinkServer {
         std::future::ready(match result {
             Ok(text) => Ok(ReadResourceResult::new(vec![
                 ResourceContents::text(text, uri).with_mime_type("application/json"),
-            ])),
+            ])
+            .into()),
             Err(err) => Err(McpError::resource_not_found(err, None)),
         })
     }
@@ -3259,26 +3550,34 @@ impl DeliberationStore {
             }
         }
 
+        let criticality = match (input.criticality, input.critical) {
+            (_, Some(true)) => AssumptionCriticality::Critical,
+            (Some(criticality), _) => criticality,
+            (None, critical) => assumption_criticality_from_text(&text, critical),
+        };
         let now = chrono::Utc::now().to_rfc3339();
         let policy = session.verification_policy.clone();
-        let status = input.status.unwrap_or_else(|| {
-            if input.evidence.is_empty() {
-                AssumptionStatus::Proposed
-            } else {
-                match evaluate_claim_status(
-                    &input.evidence,
-                    &policy,
-                    input.critical.unwrap_or(false),
-                ) {
-                    VerificationStatus::Verified => AssumptionStatus::Verified,
-                    VerificationStatus::SourceConflict => AssumptionStatus::Falsified,
-                    VerificationStatus::DataDeficit => AssumptionStatus::Unresolved,
+        let evaluated_status = evaluate_claim_status(
+            &input.evidence,
+            &policy,
+            matches!(criticality, AssumptionCriticality::Critical),
+        );
+        let status = match input.status {
+            Some(AssumptionStatus::Verified) => {
+                if matches!(evaluated_status, VerificationStatus::Verified) {
+                    AssumptionStatus::Verified
+                } else {
+                    AssumptionStatus::Unresolved
                 }
             }
-        });
-        let criticality = input
-            .criticality
-            .unwrap_or_else(|| assumption_criticality_from_text(&text, input.critical));
+            Some(status) => status,
+            None if input.evidence.is_empty() => AssumptionStatus::Proposed,
+            None => match evaluated_status {
+                VerificationStatus::Verified => AssumptionStatus::Verified,
+                VerificationStatus::SourceConflict => AssumptionStatus::Falsified,
+                VerificationStatus::DataDeficit => AssumptionStatus::Unresolved,
+            },
+        };
         let assumption_id = make_id("assumption");
         let analysis = heuristic_metadata("assumption_ledger_record", source_node_ids.clone());
 
@@ -3323,6 +3622,7 @@ impl DeliberationStore {
         input: SetAssumptionStatusInput,
     ) -> Result<SetAssumptionStatusResult, String> {
         let session = self.get_mut(&input.deliberation_id)?;
+        let policy = session.verification_policy.clone();
         let analysis = heuristic_metadata(
             "assumption_ledger_status_update",
             vec![input.assumption_id.clone()],
@@ -3332,8 +3632,6 @@ impl DeliberationStore {
             .get_mut(&input.assumption_id)
             .ok_or_else(|| format!("unknown assumption_id: {}", input.assumption_id))?;
 
-        assumption.status = input.status;
-        assumption.updated_at = chrono::Utc::now().to_rfc3339();
         if let Some(confidence) = input.confidence {
             assumption.confidence = Some(confidence.clamp(0.0, 1.0));
         }
@@ -3343,6 +3641,20 @@ impl DeliberationStore {
         if let Some(evidence) = input.evidence {
             assumption.evidence = evidence;
         }
+        assumption.status = if matches!(input.status, AssumptionStatus::Verified)
+            && !matches!(
+                evaluate_claim_status(
+                    &assumption.evidence,
+                    &policy,
+                    matches!(assumption.criticality, AssumptionCriticality::Critical),
+                ),
+                VerificationStatus::Verified
+            ) {
+            AssumptionStatus::Unresolved
+        } else {
+            input.status
+        };
+        assumption.updated_at = chrono::Utc::now().to_rfc3339();
         assumption.analysis = analysis.clone();
         let updated = assumption.clone();
         session.record_analysis(analysis);
@@ -4713,7 +5025,7 @@ impl DeliberationStore {
             .unwrap_or_else(|| session.verification_policy.clone())
             .normalize();
         session.verification_policy = policy.clone();
-        let mut matrix = Vec::new();
+        let mut updates = Vec::new();
 
         let explicit_claims = input.claims.unwrap_or_default();
         let fallback_claims = if explicit_claims.is_empty() {
@@ -4753,8 +5065,8 @@ impl DeliberationStore {
             } else {
                 ""
             };
-            matrix.push(VerificationEntry {
-                claim: claim.text,
+            updates.push(VerificationEntry {
+                claim: normalize_claim_text(&claim.text),
                 critical: claim.critical,
                 status,
                 evidence,
@@ -4766,6 +5078,19 @@ impl DeliberationStore {
             });
         }
 
+        let mut matrix = session.verification_matrix.clone();
+        for mut update in updates {
+            if let Some(existing) = matrix
+                .iter_mut()
+                .find(|entry| normalize_claim_text(&entry.claim) == update.claim)
+            {
+                update.critical = existing.critical || update.critical;
+                update.status = evaluate_claim_status(&update.evidence, &policy, update.critical);
+                *existing = update;
+            } else {
+                matrix.push(update);
+            }
+        }
         session.verification_matrix = matrix.clone();
         session.record_analysis(analysis.clone());
         let summary = VerificationSummary::from_entries(&matrix);
@@ -4787,17 +5112,22 @@ impl DeliberationStore {
     fn consensus(&mut self, input: ConsensusAnswerInput) -> Result<ConsensusAnswerResult, String> {
         let session = self.get_mut(&input.deliberation_id)?;
         let method = input.method.unwrap_or(ConsensusMethod::SelfConsistency);
-        if let Some(policy) = input.policy_override {
-            session.verification_policy = policy.normalize();
-        }
-        let analysis = heuristic_metadata("consensus_route_heuristic", session.frontier.clone());
-
+        let policy = input
+            .policy_override
+            .unwrap_or_else(|| session.verification_policy.clone())
+            .normalize();
         let mut ranked = session
             .frontier
             .iter()
             .filter_map(|id| session.nodes.get(id))
             .cloned()
             .collect::<Vec<_>>();
+        let frontier_has_agent_authored = ranked
+            .iter()
+            .any(|node| is_agent_authored(&deliberation_node_view(node)));
+        if frontier_has_agent_authored {
+            ranked.retain(|node| is_agent_authored(&deliberation_node_view(node)));
+        }
         ranked.sort_by(|a, b| {
             b.score
                 .unwrap_or(0.0)
@@ -4809,6 +5139,10 @@ impl DeliberationStore {
         if top.is_empty() {
             return Err("no frontier nodes available for consensus".to_string());
         }
+        let analysis = heuristic_metadata(
+            "consensus_route_heuristic",
+            top.iter().map(|node| node.node_id.clone()).collect(),
+        );
 
         let final_answer = top
             .iter()
@@ -4827,14 +5161,25 @@ impl DeliberationStore {
             .any(|e| e.critical && !matches!(e.status, VerificationStatus::Verified));
         let blocking_assumptions = blocking_assumption_ids(session);
         let assumption_blocked = !blocking_assumptions.is_empty();
-
-        let route_decision = if critical_unresolved || unresolved > 0 || assumption_blocked {
-            RouteDecision::GatherMoreEvidence
-        } else if matches!(session.profile, ReasoningProfile::Paranoid) {
-            RouteDecision::ProceedWithCaveats
-        } else {
-            RouteDecision::Proceed
-        };
+        let authorship_blocked = !frontier_has_agent_authored;
+        let pipeline_blocked = session
+            .stage_findings
+            .iter()
+            .any(|finding| !finding.gate_passed);
+        let checkpoint_blocked = session
+            .checkpoints
+            .last()
+            .is_some_and(|checkpoint| !checkpoint.blocking_gaps.is_empty());
+        let governance_blocked = authorship_blocked || pipeline_blocked || checkpoint_blocked;
+
+        let route_decision =
+            if critical_unresolved || unresolved > 0 || assumption_blocked || governance_blocked {
+                RouteDecision::GatherMoreEvidence
+            } else if matches!(session.profile, ReasoningProfile::Paranoid) {
+                RouteDecision::ProceedWithCaveats
+            } else {
+                RouteDecision::Proceed
+            };
         session.route_decision = Some(route_decision);
 
         let confidence =
@@ -4846,10 +5191,9 @@ impl DeliberationStore {
             blocking_assumptions.len()
         );
 
-        let policy_blocked = (critical_unresolved || assumption_blocked)
-            && session
-                .verification_policy
-                .fail_closed_on_critical_unresolved;
+        let policy_blocked = ((critical_unresolved || assumption_blocked)
+            && policy.fail_closed_on_critical_unresolved)
+            || governance_blocked;
         session.record_analysis(analysis.clone());
 
         Ok(ConsensusAnswerResult {
@@ -4872,9 +5216,6 @@ impl DeliberationStore {
         input: RunReasonKitPipelineInput,
     ) -> Result<RunReasonKitPipelineResult, String> {
         let session = self.get_mut(&input.deliberation_id)?;
-        if let Some(policy) = input.policy_override {
-            session.verification_policy = policy.normalize();
-        }
         let analysis = heuristic_metadata("reasonkit_pipeline_heuristic", session.frontier.clone());
         let stages = input.stages.unwrap_or_else(|| {
             vec![
@@ -4915,26 +5256,24 @@ impl DeliberationStore {
         };
 
         let mut pipeline_blockers = Vec::new();
-        let findings = stages
-            .iter()
-            .map(|stage| {
-                let artifact: PipelineStageArtifact = run_pipeline_stage(stage, &ctx);
-                if !artifact.gate.passed {
-                    pipeline_blockers.extend(artifact.gate.blockers.clone());
-                }
-                StageFinding {
-                    stage: artifact.stage,
-                    finding: artifact.finding,
-                    confidence: artifact.confidence,
-                    confidence_basis: artifact.confidence_basis,
-                    structured: artifact.structured,
-                    gate_passed: artifact.gate.passed,
-                    gate_blockers: artifact.gate.blockers,
-                    agent_tasks: artifact.agent_tasks,
-                    analysis: analysis.clone(),
-                }
-            })
-            .collect::<Vec<_>>();
+        let mut findings = Vec::with_capacity(stages.len());
+        for stage in &stages {
+            let artifact: PipelineStageArtifact = run_pipeline_stage(stage, &ctx)?;
+            if !artifact.gate.passed {
+                pipeline_blockers.extend(artifact.gate.blockers.clone());
+            }
+            findings.push(StageFinding {
+                stage: artifact.stage,
+                finding: artifact.finding,
+                confidence: artifact.confidence,
+                confidence_basis: artifact.confidence_basis,
+                structured: artifact.structured,
+                gate_passed: artifact.gate.passed,
+                gate_blockers: artifact.gate.blockers,
+                agent_tasks: artifact.agent_tasks,
+                analysis: analysis.clone(),
+            });
+        }
 
         let has_critical_unresolved = session.verification_matrix.iter().any(|e| {
             e.critical
@@ -5007,9 +5346,10 @@ impl DeliberationStore {
             .get(&input.deliberation_id)
             .ok_or_else(|| "unknown deliberation_id".to_string())?;
         let include_raw = input.include_raw_thoughts.unwrap_or(false);
+        let audit_id = make_id("audit");
 
         let mut payload = json!({
-            "audit_id": make_id("audit"),
+            "audit_id": audit_id,
             "deliberation_id": session.deliberation_id,
             "session_id": session.session_id,
             "mode": session.mode,
@@ -5040,10 +5380,20 @@ impl DeliberationStore {
             payload["nodes"] = json!(session.nodes);
         }
 
+        let artifact = match input.format {
+            AuditFormat::Json => render_canonical_json(&payload)?,
+            AuditFormat::Markdown => {
+                render_audit_markdown(&audit_id, session, include_raw, &payload)
+            }
+        };
+
         Ok(ExportReasoningAuditResult {
             deliberation_id: session.deliberation_id.clone(),
-            audit_id: make_id("audit"),
+            audit_id,
             payload,
+            format: input.format,
+            content_type: input.format.content_type().to_string(),
+            artifact,
         })
     }
 
@@ -5321,6 +5671,15 @@ impl DeliberationStore {
         &mut self,
         input: RecordReasoningActionInput,
     ) -> Result<RecordReasoningActionResult, String> {
+        if input
+            .tags
+            .as_ref()
+            .is_some_and(|tags| tags.iter().any(|tag| tag == "agent-authored"))
+        {
+            return Err(
+                "invalid action tag: `agent-authored` is reserved for add_thought_node".to_string(),
+            );
+        }
         let session = self.get_mut(&input.deliberation_id)?;
         if session.nodes.len() >= session.limits.max_nodes as usize {
             return Err("capacity limit reached: max_nodes exceeded".to_string());
@@ -5384,8 +5743,15 @@ impl DeliberationStore {
         input: RecordReasoningObservationInput,
     ) -> Result<RecordReasoningObservationResult, String> {
         let session = self.get_mut(&input.deliberation_id)?;
-        if !session.nodes.contains_key(&input.action_node_id) {
-            return Err(format!("unknown action_node_id: {}", input.action_node_id));
+        match session.nodes.get(&input.action_node_id) {
+            None => return Err(format!("unknown action_node_id: {}", input.action_node_id)),
+            Some(node) if !matches!(node.node_type, ThoughtNodeType::Action { .. }) => {
+                return Err(format!(
+                    "invalid action_node_id; must reference an Action node: {}",
+                    input.action_node_id
+                ));
+            }
+            Some(_) => {}
         }
         if session.nodes.len() >= session.limits.max_nodes as usize {
             return Err("capacity limit reached: max_nodes exceeded".to_string());
@@ -5416,6 +5782,22 @@ impl DeliberationStore {
                 result: result.clone(),
             },
         };
+        let action_node = session
+            .nodes
+            .get_mut(&input.action_node_id)
+            .expect("validated action node must remain present");
+        if let ThoughtNodeType::Action {
+            tool_name,
+            parameters,
+            status,
+        } = &mut action_node.node_type
+        {
+            *status = ActionStatus::Complete;
+            action_node.content = format!(
+                "[ACTION] tool={} status={:?}\nparams={}",
+                tool_name, status, parameters
+            );
+        }
         session.insert_node(node);
         session.record_analysis(analysis.clone());
         Ok(RecordReasoningObservationResult {
@@ -7159,74 +7541,353 @@ fn evaluate_claim_status(
     }
 }
 
-fn has_any_ci<S: AsRef<str>>(text: &str, needles: &[S]) -> bool {
-    needles
-        .iter()
-        .map(|n| n.as_ref().to_lowercase())
-        .any(|n| text.contains(&n))
-}
-
-fn has_any_ci_groups(text: &str, groups: &[&[String]]) -> bool {
-    groups.iter().any(|group| has_any_ci(text, group))
-}
-
-fn dedup_in_place(values: &mut Vec<String>) {
-    let mut seen = HashSet::new();
-    values.retain(|v| seen.insert(v.clone()));
+fn normalize_claim_text(text: &str) -> String {
+    text.split_whitespace().collect::<Vec<_>>().join(" ")
 }
 
-fn normalize_alias_list(values: Vec<String>) -> Vec<String> {
-    let mut out = values
-        .into_iter()
-        .map(|v| v.trim().to_lowercase())
-        .filter(|v| !v.is_empty())
-        .collect::<Vec<_>>();
-    dedup_in_place(&mut out);
-    out
+fn render_canonical_json(payload: &Value) -> Result<String, String> {
+    serde_json::to_string_pretty(&canonicalize_json(payload))
+        .map_err(|err| format!("audit JSON rendering failed: {err}"))
 }
 
-fn extend_unique(dest: &mut Vec<String>, add: &[String]) {
-    let mut set = dest.iter().cloned().collect::<HashSet<_>>();
-    for value in add {
-        let value = value.trim().to_lowercase();
-        if !value.is_empty() && set.insert(value.clone()) {
-            dest.push(value);
+fn canonicalize_json(value: &Value) -> Value {
+    match value {
+        Value::Object(map) => {
+            let mut keys = map.keys().collect::<Vec<_>>();
+            keys.sort();
+            let mut sorted = Map::new();
+            for key in keys {
+                sorted.insert(key.clone(), canonicalize_json(&map[key]));
+            }
+            Value::Object(sorted)
         }
+        Value::Array(items) => Value::Array(items.iter().map(canonicalize_json).collect()),
+        _ => value.clone(),
     }
 }
 
-fn explain_tool(tool: &str) -> String {
-    match tool {
-        "reasoning_autopilot" => {
-            "Auto mode: route intent and run the default audited reasoning flow.".to_string()
-        }
-        "run_thinking_mode" => {
-            "Execute Auto, Quick, Explore, Map, Sketch, or Test as one audited workflow."
-                .to_string()
+fn render_audit_markdown(
+    audit_id: &str,
+    session: &DeliberationSession,
+    include_raw_thoughts: bool,
+    payload: &Value,
+) -> String {
+    let route = session
+        .route_decision
+        .as_ref()
+        .map(audit_label)
+        .unwrap_or_else(|| "not_recorded".to_string());
+    let active_mode = payload
+        .pointer("/provider_capabilities/active_mode")
+        .and_then(Value::as_str)
+        .unwrap_or("unknown");
+    let semantic_engine = payload
+        .pointer("/provider_capabilities/semantic_engine")
+        .and_then(Value::as_str)
+        .unwrap_or("unknown");
+    let mut providers = Vec::new();
+    for analysis in &session.analysis_runs {
+        if let Some(provider) = &analysis.provider {
+            providers.push(provider.clone());
         }
-        "reasoning_intent_router" => {
-            "Translate normal wording into Thinking Modes and MCP tool calls.".to_string()
+    }
+    for entry in &session.verification_matrix {
+        if let Some(provider) = &entry.analysis.provider {
+            providers.push(provider.clone());
         }
-        "sequentialthinking_tools" => "Quick linear thought capture (CoT baseline).".to_string(),
-        "get_thinking_history" => "Inspect current quick-thought assumptions.".to_string(),
-        "start_deliberation" => "Open structured deep reasoning session.".to_string(),
-        "expand_thoughts" => "Create heuristic alternatives/branches from frontier.".to_string(),
-        "start_skeleton_of_thought" => {
-            "Create a routed Skeleton-of-Thought phase with scaffold skeleton nodes.".to_string()
+    }
+    for assumption in session.assumptions.values() {
+        if let Some(provider) = &assumption.analysis.provider {
+            providers.push(provider.clone());
         }
-        "run_skeleton_elaboration" => {
-            "Elaborate SoT skeletons into prompt packs or bounded scaffold nodes.".to_string()
+    }
+    for stage in &session.stage_findings {
+        if let Some(provider) = &stage.analysis.provider {
+            providers.push(provider.clone());
         }
-        "score_thoughts" => "Rank candidates by weighted heuristic proxies.".to_string(),
-        "prune_thoughts" => "Keep best/diverse candidates, remove weak branches.".to_string(),
-        "link_thoughts" => "Add typed GoT edges between existing thought nodes.".to_string(),
-        "merge_thought_branches" => {
-            "Converge multiple branches into one merged synthesis node.".to_string()
+    }
+    for node in session.nodes.values() {
+        if let Some(provider) = &node.provenance.provider {
+            providers.push(provider.clone());
         }
-        "distill_thought_cluster" => {
-            "Compress multiple nodes into a bullet distill with dependency edges.".to_string()
+    }
+    providers.sort();
+    providers.dedup();
+    let providers = if providers.is_empty() {
+        "none recorded".to_string()
+    } else {
+        providers.join(", ")
+    };
+
+    let mut out = String::from("# Reasoning Audit\n\n");
+    out.push_str("## Audit metadata\n\n");
+    out.push_str("| Field | Value |\n| --- | --- |\n");
+    push_markdown_row(&mut out, &["Audit ID", audit_id]);
+    push_markdown_row(&mut out, &["Deliberation ID", &session.deliberation_id]);
+    push_markdown_row(&mut out, &["Session", &session.session_id]);
+    push_markdown_row(&mut out, &["Mode", &audit_label(&session.mode)]);
+    push_markdown_row(&mut out, &["Profile", &audit_label(&session.profile)]);
+    push_markdown_row(&mut out, &["Goal", &session.goal]);
+    push_markdown_row(&mut out, &["Route decision", &route]);
+
+    out.push_str("\n## Provenance and provider boundaries\n\n");
+    if active_mode == "heuristic" {
+        out.push_str(
+            "> **HEURISTIC OUTPUT:** ReasonKit supplied deterministic structural analysis; no model-backed semantic verification is implied.\n\n",
+        );
+    } else {
+        out.push_str("> **PROVIDER-BACKED OUTPUT:** Review the provider metadata below.\n\n");
+    }
+    out.push_str("| Field | Value |\n| --- | --- |\n");
+    push_markdown_row(&mut out, &["Active analysis mode", active_mode]);
+    push_markdown_row(&mut out, &["Semantic engine", semantic_engine]);
+    push_markdown_row(&mut out, &["Recorded providers", &providers]);
+    push_markdown_row(
+        &mut out,
+        &[
+            "Boundary",
+            "Model- and retrieval-backed claims remain fail-closed unless a real provider binding is active; canonical provider details remain in payload.provider_capabilities.",
+        ],
+    );
+
+    out.push_str("\n## Verification matrix\n\n");
+    out.push_str("| Claim | Critical | Status | Evidence sources |\n| --- | --- | --- | --- |\n");
+    let mut verification = session.verification_matrix.iter().collect::<Vec<_>>();
+    verification.sort_by(|left, right| left.claim.cmp(&right.claim));
+    if verification.is_empty() {
+        push_markdown_row(&mut out, &["none", "no", "not_recorded", "none"]);
+    } else {
+        for entry in verification {
+            let mut sources = entry
+                .evidence
+                .iter()
+                .map(|evidence| format!("{} ({})", evidence.source, audit_label(&evidence.tier)))
+                .collect::<Vec<_>>();
+            sources.sort();
+            sources.dedup();
+            let sources = if sources.is_empty() {
+                "none".to_string()
+            } else {
+                sources.join(", ")
+            };
+            push_markdown_row(
+                &mut out,
+                &[
+                    &entry.claim,
+                    if entry.critical { "yes" } else { "no" },
+                    &audit_label(&entry.status),
+                    &sources,
+                ],
+            );
         }
-        "refine_thoughts" => {
+    }
+
+    out.push_str("\n## Assumptions\n\n");
+    out.push_str("| ID | Assumption | Status | Criticality |\n| --- | --- | --- | --- |\n");
+    let mut assumptions = session.assumptions.values().collect::<Vec<_>>();
+    assumptions.sort_by(|left, right| left.assumption_id.cmp(&right.assumption_id));
+    if assumptions.is_empty() {
+        push_markdown_row(&mut out, &["none", "none", "not_recorded", "none"]);
+    } else {
+        for assumption in assumptions {
+            push_markdown_row(
+                &mut out,
+                &[
+                    &assumption.assumption_id,
+                    &assumption.text,
+                    &audit_label(&assumption.status),
+                    &audit_label(&assumption.criticality),
+                ],
+            );
+        }
+    }
+
+    out.push_str("\n## Quality checkpoints\n\n");
+    out.push_str("| ID | Label | Route | Blocking gaps |\n| --- | --- | --- | --- |\n");
+    let mut checkpoints = session.checkpoints.iter().collect::<Vec<_>>();
+    checkpoints.sort_by(|left, right| left.checkpoint_id.cmp(&right.checkpoint_id));
+    if checkpoints.is_empty() {
+        push_markdown_row(&mut out, &["none", "none", "not_recorded", "none"]);
+    } else {
+        for checkpoint in checkpoints {
+            push_markdown_row(
+                &mut out,
+                &[
+                    &checkpoint.checkpoint_id,
+                    &checkpoint.label,
+                    &audit_label(&checkpoint.route_decision),
+                    &sorted_markdown_values(&checkpoint.blocking_gaps),
+                ],
+            );
+        }
+    }
+
+    out.push_str("\n## Pipeline gates\n\n");
+    out.push_str("| Stage | Gate | Blockers |\n| --- | --- | --- |\n");
+    let mut stages = session.stage_findings.iter().collect::<Vec<_>>();
+    stages.sort_by(|left, right| left.stage.cmp(&right.stage));
+    if stages.is_empty() {
+        push_markdown_row(&mut out, &["none", "not_recorded", "none"]);
+    } else {
+        for stage in stages {
+            push_markdown_row(
+                &mut out,
+                &[
+                    &stage.stage,
+                    if stage.gate_passed {
+                        "passed"
+                    } else {
+                        "blocked"
+                    },
+                    &sorted_markdown_values(&stage.gate_blockers),
+                ],
+            );
+        }
+    }
+
+    out.push_str("\n## Raw thoughts\n\n");
+    if include_raw_thoughts {
+        out.push_str("| Node ID | Branch | Provider | Content |\n| --- | --- | --- | --- |\n");
+        let mut nodes = session.nodes.values().collect::<Vec<_>>();
+        nodes.sort_by(|left, right| left.node_id.cmp(&right.node_id));
+        for node in nodes {
+            push_markdown_row(
+                &mut out,
+                &[
+                    &node.node_id,
+                    node.branch_id.as_deref().unwrap_or("none"),
+                    node.provenance.provider.as_deref().unwrap_or("none"),
+                    &node.content,
+                ],
+            );
+        }
+    } else {
+        out.push_str(
+            "Raw node content omitted. Re-export with `include_raw_thoughts=true` only when disclosure is appropriate.\n",
+        );
+    }
+
+    out
+}
+
+fn audit_label<T: Serialize>(value: &T) -> String {
+    serde_json::to_value(value)
+        .ok()
+        .and_then(|value| value.as_str().map(str::to_string))
+        .unwrap_or_else(|| "unknown".to_string())
+}
+
+fn sorted_markdown_values(values: &[String]) -> String {
+    if values.is_empty() {
+        return "none".to_string();
+    }
+    let mut values = values.to_vec();
+    values.sort();
+    values.dedup();
+    values.join("\n")
+}
+
+fn push_markdown_row(out: &mut String, cells: &[&str]) {
+    out.push_str("| ");
+    out.push_str(
+        &cells
+            .iter()
+            .map(|cell| markdown_table_cell(cell))
+            .collect::<Vec<_>>()
+            .join(" | "),
+    );
+    out.push_str(" |\n");
+}
+
+fn markdown_table_cell(value: &str) -> String {
+    value
+        .replace("\r\n", "\n")
+        .replace('\r', "\n")
+        .replace('\\', "\\\\")
+        .replace('&', "&amp;")
+        .replace('<', "&lt;")
+        .replace('>', "&gt;")
+        .replace('|', "\\|")
+        .replace('!', "&#33;")
+        .replace('[', "&#91;")
+        .replace(']', "&#93;")
+        .replace('(', "&#40;")
+        .replace(')', "&#41;")
+        .replace('*', "&#42;")
+        .replace('_', "&#95;")
+        .replace('`', "&#96;")
+        .replace('~', "&#126;")
+        .replace('\n', "<br>")
+}
+
+fn has_any_ci<S: AsRef<str>>(text: &str, needles: &[S]) -> bool {
+    needles
+        .iter()
+        .map(|n| n.as_ref().to_lowercase())
+        .any(|n| text.contains(&n))
+}
+
+fn has_any_ci_groups(text: &str, groups: &[&[String]]) -> bool {
+    groups.iter().any(|group| has_any_ci(text, group))
+}
+
+fn dedup_in_place(values: &mut Vec<String>) {
+    let mut seen = HashSet::new();
+    values.retain(|v| seen.insert(v.clone()));
+}
+
+fn normalize_alias_list(values: Vec<String>) -> Vec<String> {
+    let mut out = values
+        .into_iter()
+        .map(|v| v.trim().to_lowercase())
+        .filter(|v| !v.is_empty())
+        .collect::<Vec<_>>();
+    dedup_in_place(&mut out);
+    out
+}
+
+fn extend_unique(dest: &mut Vec<String>, add: &[String]) {
+    let mut set = dest.iter().cloned().collect::<HashSet<_>>();
+    for value in add {
+        let value = value.trim().to_lowercase();
+        if !value.is_empty() && set.insert(value.clone()) {
+            dest.push(value);
+        }
+    }
+}
+
+fn explain_tool(tool: &str) -> String {
+    match tool {
+        "reasoning_autopilot" => {
+            "Auto mode: route intent and run the default audited reasoning flow.".to_string()
+        }
+        "run_thinking_mode" => {
+            "Execute Auto, Quick, Explore, Map, Sketch, or Test as one audited workflow."
+                .to_string()
+        }
+        "reasoning_intent_router" => {
+            "Translate normal wording into Thinking Modes and MCP tool calls.".to_string()
+        }
+        "sequentialthinking_tools" => "Quick linear thought capture (CoT baseline).".to_string(),
+        "get_thinking_history" => "Inspect current quick-thought assumptions.".to_string(),
+        "start_deliberation" => "Open structured deep reasoning session.".to_string(),
+        "expand_thoughts" => "Create heuristic alternatives/branches from frontier.".to_string(),
+        "start_skeleton_of_thought" => {
+            "Create a routed Skeleton-of-Thought phase with scaffold skeleton nodes.".to_string()
+        }
+        "run_skeleton_elaboration" => {
+            "Elaborate SoT skeletons into prompt packs or bounded scaffold nodes.".to_string()
+        }
+        "score_thoughts" => "Rank candidates by weighted heuristic proxies.".to_string(),
+        "prune_thoughts" => "Keep best/diverse candidates, remove weak branches.".to_string(),
+        "link_thoughts" => "Add typed GoT edges between existing thought nodes.".to_string(),
+        "merge_thought_branches" => {
+            "Converge multiple branches into one merged synthesis node.".to_string()
+        }
+        "distill_thought_cluster" => {
+            "Compress multiple nodes into a bullet distill with dependency edges.".to_string()
+        }
+        "refine_thoughts" => {
             "Create an explicit feedback-loop correction node with a refines edge.".to_string()
         }
         "record_assumption" => "Add an explicit premise to the assumption ledger.".to_string(),
@@ -7854,17 +8515,38 @@ fn select_thinking_mode(inferred_modes: &[String], aliases: &ReasoningAliases) -
     }
 }
 
-fn thinking_modes_resource(aliases: &ReasoningAliases) -> Value {
+fn thinking_modes_resource(aliases: &ReasoningAliases, tool_pack: ToolPack) -> Value {
     let modes = all_thinking_modes()
         .into_iter()
-        .map(|mode| thinking_mode_contract(mode, aliases))
+        .map(|mode| {
+            let mut contract = thinking_mode_contract(mode, aliases);
+            contract
+                .primary_tools
+                .retain(|tool| tool_pack.contains(tool));
+            contract.compatible_internal_values.retain(|value| {
+                ToolPack::minimum_for(value).is_none() || tool_pack.contains(value)
+            });
+            contract
+        })
         .collect::<Vec<_>>();
+    let compatible_tools = |tools: &[&str]| -> Vec<String> {
+        tools
+            .iter()
+            .copied()
+            .filter(|tool| ToolPack::minimum_for(tool).is_none() || tool_pack.contains(tool))
+            .map(str::to_string)
+            .collect()
+    };
     json!({
         "uri": "reasoning://thinking-modes",
         "version": 1,
         "default_mode": ThinkingMode::Auto,
         "default_tool": "run_thinking_mode",
-        "compatibility_default_tool": "reasoning_autopilot",
+        "active_tool_pack": tool_pack.as_str(),
+        "tool_pack_resource": "reasoning://config/tool-packs",
+        "compatibility_default_tool": tool_pack
+            .contains("reasoning_autopilot")
+            .then_some("reasoning_autopilot"),
         "mode_order": all_thinking_modes()
             .into_iter()
             .map(thinking_mode_id)
@@ -7874,9 +8556,9 @@ fn thinking_modes_resource(aliases: &ReasoningAliases) -> Value {
             "quick": ["cot"],
             "explore": ["tot"],
             "map": ["got"],
-            "sketch": ["got", "start_skeleton_of_thought", "run_skeleton_elaboration"],
-            "test": ["reasonkit", "verify_thoughts", "run_reasonkit_pipeline"],
-            "auto": ["reasoning_autopilot", "reasoning_intent_router"]
+            "sketch": compatible_tools(&["got", "start_skeleton_of_thought", "run_skeleton_elaboration"]),
+            "test": compatible_tools(&["reasonkit", "verify_thoughts", "run_reasonkit_pipeline"]),
+            "auto": compatible_tools(&["reasoning_autopilot", "reasoning_intent_router"])
         },
         "resource_contract": {
             "resources_are_uri_addressed": true,
@@ -9026,13 +9708,49 @@ fn parse_int_env(name: &str, fallback: usize) -> usize {
 }
 
 fn json_tool_success<T: Serialize>(data: &T) -> CallToolResult {
-    let text = serde_json::to_string_pretty(data).unwrap_or_else(|_| "{}".to_string());
-    CallToolResult::success(vec![Content::text(text)])
+    json_tool_result(data, false)
 }
 
 fn json_tool_error<T: Serialize>(data: &T) -> CallToolResult {
-    let text = serde_json::to_string_pretty(data).unwrap_or_else(|_| "{}".to_string());
-    CallToolResult::error(vec![Content::text(text)])
+    json_tool_result(data, true)
+}
+
+fn consensus_blocked_tool_result(result: &ConsensusAnswerResult) -> CallToolResult {
+    json_tool_error(&json!({
+        "code": "consensus_blocked",
+        "error": "consensus is blocked by unresolved governance gates",
+        "retryable": false,
+        "deliberation_id": result.deliberation_id,
+        "route_decision": result.route_decision,
+        "policy_blocked": true,
+        "recommended_action": "resolve the reported evidence, assumption, authorship, pipeline, or checkpoint blockers before requesting consensus again",
+    }))
+}
+
+fn json_tool_result<T: Serialize>(data: &T, is_error: bool) -> CallToolResult {
+    match serde_json::to_value(data) {
+        Ok(value) => json_value_tool_result(value, is_error),
+        Err(_) => json_value_tool_result(
+            json!({
+                "code": "serialization_error",
+                "message": "failed to serialize tool result",
+                "retryable": false,
+            }),
+            true,
+        ),
+    }
+}
+
+fn json_value_tool_result(value: Value, is_error: bool) -> CallToolResult {
+    let text = serde_json::to_string_pretty(&value)
+        .expect("serde_json::Value must always serialize to JSON text");
+    let mut result = if is_error {
+        CallToolResult::structured_error(value)
+    } else {
+        CallToolResult::structured(value)
+    };
+    result.content = vec![ContentBlock::text(text)];
+    result
 }
 
 fn canonical_token(raw: &str) -> String {
@@ -11292,6 +12010,37 @@ fn provider_capabilities() -> Value {
     })
 }
 
+fn tool_pack_resource(active: ToolPack) -> Value {
+    json!({
+        "uri": "reasoning://config/tool-packs",
+        "version": 1,
+        "default": ToolPack::Core.as_str(),
+        "active": active.as_str(),
+        "active_tool_count": active.members().len(),
+        "restart_required": true,
+        "environment_variable": "REASONKIT_TOOL_PACK",
+        "invalid_value_behavior": "warn on stderr and use the safe core fallback",
+        "compatibility_rollback": "REASONKIT_TOOL_PACK=full",
+        "packs": {
+            "core": {
+                "tool_count": ToolPack::Core.members().len(),
+                "purpose": "agent-native governance golden path",
+                "tools": ToolPack::Core.members(),
+            },
+            "standard": {
+                "tool_count": ToolPack::Standard.members().len(),
+                "purpose": "core plus everyday Explore, Map, Sketch, replay, and tuning operators",
+                "tools": ToolPack::Standard.members(),
+            },
+            "full": {
+                "tool_count": ToolPack::Full.members().len(),
+                "purpose": "compatibility surface including legacy, specialist, web-sidecar, and experimental memory tools",
+                "tools": ToolPack::Full.members(),
+            }
+        }
+    })
+}
+
 fn transport_capabilities() -> Value {
     json!({
         "active_transport": "stdio",
@@ -11830,24 +12579,6 @@ struct ThinkQueryInput {
     min_score: Option<f32>,
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
-struct HotMemoryEntryResult {
-    id: String,
-    content: String,
-    score: f32,
-    metadata: serde_json::Value,
-}
-
-#[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
-struct ThinkQueryResult {
-    results: Vec<HotMemoryEntryResult>,
-    latency_ms: u128,
-    /// Cross-session failure-mode hints (host agent protocol; not LLM-generated).
-    #[serde(default)]
-    failure_coaching: Vec<String>,
-    semantic_engine: String,
-}
-
 #[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
 #[schemars(transform = vertex_compat_schema)]
 struct RunThinkingModeInput {
@@ -12567,6 +13298,7 @@ fn deliberation_node_view(node: &DeliberationNode) -> NodeView {
         score: node.score,
         branch_id: node.branch_id.clone(),
         node_type: node.node_type.clone(),
+        agent_authored_provenance: matches!(node.provenance.mode, AnalysisMode::AgentAuthored),
     }
 }
 
@@ -14046,11 +14778,31 @@ struct RunReasonKitPipelineResult {
     analysis: AnalysisMetadata,
 }
 
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[schemars(inline)]
+#[serde(rename_all = "lowercase")]
+enum AuditFormat {
+    #[default]
+    Json,
+    Markdown,
+}
+
+impl AuditFormat {
+    fn content_type(self) -> &'static str {
+        match self {
+            Self::Json => "application/json",
+            Self::Markdown => "text/markdown; charset=utf-8",
+        }
+    }
+}
+
 #[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
 #[schemars(transform = vertex_compat_schema)]
 struct ExportReasoningAuditInput {
     deliberation_id: String,
     #[serde(default)]
+    format: AuditFormat,
+    #[serde(default)]
     #[serde(deserialize_with = "deserialize_opt_bool")]
     include_raw_thoughts: Option<bool>,
 }
@@ -14181,6 +14933,9 @@ struct ExportReasoningAuditResult {
     deliberation_id: String,
     audit_id: String,
     payload: Value,
+    format: AuditFormat,
+    content_type: String,
+    artifact: String,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize, JsonSchema)]
@@ -14400,6 +15155,7 @@ fn reasoning_resource_uris() -> Vec<&'static str> {
         "reasoning://session/{id}/route-decision",
         "reasoning://config/aliases",
         "reasoning://config/providers",
+        "reasoning://config/tool-packs",
         "reasoning://config/web-research",
         "reasoning://config/transports",
         "reasoning://workflows/feature-triage",
@@ -14430,6 +15186,12 @@ fn native_reasoning_resources() -> Vec<Resource> {
             "Provider Capabilities",
             "Configured analysis provider boundaries and limitations.",
         ),
+        (
+            "reasoning://config/tool-packs",
+            "tool-packs",
+            "Tool Packs",
+            "Active progressive tool surface and exact core, standard, and full memberships.",
+        ),
         (
             "reasoning://config/web-research",
             "web-research-config",
@@ -14469,13 +15231,10 @@ fn native_reasoning_resources() -> Vec<Resource> {
     ]
     .into_iter()
     .map(|(uri, name, title, description)| {
-        Annotated::new(
-            RawResource::new(uri, name)
-                .with_title(title)
-                .with_description(description)
-                .with_mime_type("application/json"),
-            None,
-        )
+        Resource::new(uri, name)
+            .with_title(title)
+            .with_description(description)
+            .with_mime_type("application/json")
     })
     .collect()
 }
@@ -14539,20 +15298,25 @@ fn native_reasoning_resource_templates() -> Vec<ResourceTemplate> {
     ]
     .into_iter()
     .map(|(uri_template, name, title, description)| {
-        Annotated::new(
-            RawResourceTemplate::new(uri_template, name)
-                .with_title(title)
-                .with_description(description)
-                .with_mime_type("application/json"),
-            None,
-        )
+        ResourceTemplate::new(uri_template, name)
+            .with_title(title)
+            .with_description(description)
+            .with_mime_type("application/json")
     })
     .collect()
 }
 
-fn read_reasoning_resource_payload(store: &DeliberationStore, uri: &str) -> Result<Value, String> {
+fn read_reasoning_resource_payload(
+    store: &DeliberationStore,
+    uri: &str,
+    tool_pack: ToolPack,
+) -> Result<Value, String> {
     if uri == "reasoning://thinking-modes" {
-        return Ok(thinking_modes_resource(&store.aliases));
+        return Ok(thinking_modes_resource(&store.aliases, tool_pack));
+    }
+
+    if uri == "reasoning://config/tool-packs" {
+        return Ok(tool_pack_resource(tool_pack));
     }
 
     if uri == "reasoning://config/aliases" {
@@ -14819,6 +15583,7 @@ mod input_compatibility_tests {
         }))
         .expect("audit export should accept boolean-like strings");
         assert_eq!(audit.include_raw_thoughts, Some(true));
+        assert_eq!(audit.format, AuditFormat::Json);
 
         let sot_elaboration: RunSkeletonElaborationInput = serde_json::from_value(json!({
             "deliberation_id": "delib-test",
@@ -14996,6 +15761,7 @@ mod input_compatibility_tests {
         let mode_catalog = read_reasoning_resource_payload(
             &DeliberationStore::new(64),
             "reasoning://thinking-modes",
+            ToolPack::Full,
         )
         .expect("thinking modes resource");
         assert_eq!(mode_catalog["default_mode"], json!("auto"));
@@ -15004,6 +15770,7 @@ mod input_compatibility_tests {
         let feature_workflow = read_reasoning_resource_payload(
             &DeliberationStore::new(64),
             "reasoning://workflows/feature-triage",
+            ToolPack::Full,
         )
         .expect("feature triage resource");
         assert_eq!(
@@ -15326,6 +16093,172 @@ mod architecture_contract_tests {
     use super::*;
     use std::panic::AssertUnwindSafe;
 
+    fn qualifying_tier1_evidence(source: &str) -> EvidenceItem {
+        EvidenceItem {
+            source: source.to_string(),
+            tier: EvidenceTier::Tier1,
+            independence_group: source.to_string(),
+            supports: true,
+            contradictory: false,
+            unambiguous: true,
+        }
+    }
+
+    fn seeded_audit_store() -> (DeliberationStore, String) {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("audit-session".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Review | release\nwithout <unsafe> claims".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Paranoid),
+            limits: None,
+            verification_policy: None,
+        });
+        let analysis = AnalysisMetadata {
+            mode: AnalysisMode::Heuristic,
+            method: "audit_fixture".to_string(),
+            provider: Some("reasonkit-think/heuristic".to_string()),
+            model: None,
+            generated_at: "2026-08-23T00:00:00Z".to_string(),
+            source_node_ids: Vec::new(),
+            limitations: heuristic_limitations(),
+        };
+        let session = store
+            .deliberations
+            .get_mut(&started.deliberation_id)
+            .expect("audit fixture session");
+        session.route_decision = Some(RouteDecision::GatherMoreEvidence);
+        session.verification_matrix = vec![
+            VerificationEntry {
+                claim: "Zeta claim".to_string(),
+                critical: false,
+                status: VerificationStatus::DataDeficit,
+                evidence: vec![EvidenceItem {
+                    source: "z-source".to_string(),
+                    tier: EvidenceTier::Tier2,
+                    independence_group: "z".to_string(),
+                    supports: true,
+                    contradictory: false,
+                    unambiguous: false,
+                }],
+                notes: "needs another source".to_string(),
+                analysis: analysis.clone(),
+            },
+            VerificationEntry {
+                claim: "Alpha | claim\nnext <tag>".to_string(),
+                critical: true,
+                status: VerificationStatus::Verified,
+                evidence: vec![
+                    qualifying_tier1_evidence("z-source"),
+                    qualifying_tier1_evidence("a|source"),
+                ],
+                notes: "verified".to_string(),
+                analysis: analysis.clone(),
+            },
+        ];
+        for (id, text, status, criticality) in [
+            (
+                "assumption-z",
+                "Zeta assumption",
+                AssumptionStatus::Proposed,
+                AssumptionCriticality::Low,
+            ),
+            (
+                "assumption-a",
+                "Alpha | assumption\nwith <tag>",
+                AssumptionStatus::Unresolved,
+                AssumptionCriticality::Critical,
+            ),
+        ] {
+            session.assumptions.insert(
+                id.to_string(),
+                AssumptionRecord {
+                    assumption_id: id.to_string(),
+                    text: text.to_string(),
+                    confidence: None,
+                    criticality,
+                    status,
+                    verifiable: true,
+                    depends_on: Vec::new(),
+                    invalidates: Vec::new(),
+                    source_node_ids: Vec::new(),
+                    evidence: Vec::new(),
+                    created_at: "2026-08-23T00:00:00Z".to_string(),
+                    updated_at: "2026-08-23T00:00:00Z".to_string(),
+                    notes: None,
+                    analysis: analysis.clone(),
+                },
+            );
+        }
+        session.checkpoints = vec![
+            QualityCheckpoint {
+                checkpoint_id: "checkpoint-z".to_string(),
+                label: "Zeta checkpoint".to_string(),
+                created_at: "2026-08-23T00:00:00Z".to_string(),
+                quality_score: 0.4,
+                quality_tier: QualityTier::Poor,
+                route_decision: RouteDecision::GatherMoreEvidence,
+                blocking_gaps: vec!["z checkpoint blocker".to_string()],
+                recommended_tools: vec!["verify_thoughts".to_string()],
+            },
+            QualityCheckpoint {
+                checkpoint_id: "checkpoint-a".to_string(),
+                label: "Alpha | checkpoint".to_string(),
+                created_at: "2026-08-23T00:00:00Z".to_string(),
+                quality_score: 0.5,
+                quality_tier: QualityTier::Adequate,
+                route_decision: RouteDecision::DeferToHuman,
+                blocking_gaps: vec!["a | checkpoint blocker".to_string()],
+                recommended_tools: vec!["consensus_answer".to_string()],
+            },
+        ];
+        session.stage_findings = vec![
+            StageFinding {
+                stage: "z-stage".to_string(),
+                finding: "z finding".to_string(),
+                confidence: 0.4,
+                confidence_basis: "fixture".to_string(),
+                structured: json!({}),
+                gate_passed: false,
+                gate_blockers: vec!["z pipeline blocker".to_string()],
+                agent_tasks: Vec::new(),
+                analysis: analysis.clone(),
+            },
+            StageFinding {
+                stage: "a-stage".to_string(),
+                finding: "a finding".to_string(),
+                confidence: 0.7,
+                confidence_basis: "fixture".to_string(),
+                structured: json!({}),
+                gate_passed: false,
+                gate_blockers: vec!["a | pipeline blocker".to_string()],
+                agent_tasks: Vec::new(),
+                analysis: analysis.clone(),
+            },
+        ];
+        for (id, content) in [
+            ("node-z", "RAW_ZETA_NODE_CONTENT"),
+            (
+                "node-a",
+                "RAW_ALPHA_NODE_CONTENT | line\n<script>alert(1)</script>",
+            ),
+        ] {
+            session.insert_node(DeliberationNode::thought(
+                id.to_string(),
+                Vec::new(),
+                content.to_string(),
+                None,
+                "2026-08-23T00:00:00Z".to_string(),
+                None,
+                vec!["fixture".to_string()],
+                analysis.clone(),
+            ));
+        }
+
+        (store, started.deliberation_id)
+    }
+
     #[test]
     fn heuristic_outputs_are_labeled_with_provenance_and_no_semantic_confidence() {
         let mut store = DeliberationStore::new(64);
@@ -15403,6 +16336,7 @@ mod architecture_contract_tests {
         let audit = store
             .export_audit(ExportReasoningAuditInput {
                 deliberation_id: started.deliberation_id,
+                format: AuditFormat::Json,
                 include_raw_thoughts: Some(true),
             })
             .expect("audit should work");
@@ -15419,43 +16353,842 @@ mod architecture_contract_tests {
     }
 
     #[test]
-    fn graph_indexes_track_lineage_and_strict_cycle_rejections() {
+    fn verification_upserts_normalized_claim_and_preserves_unrelated_blockers() {
         let mut store = DeliberationStore::new(64);
         let started = store.start(StartDeliberationInput {
-            session_id: Some("graph-indexes".to_string()),
-            mode: Some(ReasoningMode::Got),
-            goal: "Maintain graph indexes".to_string(),
+            session_id: Some("verification-upsert".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Preserve verification history".to_string(),
             constraints: None,
-            profile: Some(ReasoningProfile::Balanced),
+            profile: Some(ReasoningProfile::Paranoid),
             limits: None,
             verification_policy: None,
         });
-        let root_id = store
-            .deliberations
-            .get(&started.deliberation_id)
-            .and_then(|session| session.frontier.first())
-            .cloned()
-            .expect("root frontier");
 
-        let expanded = store
-            .expand(ExpandThoughtsInput {
+        store
+            .verify(VerifyThoughtsInput {
                 deliberation_id: started.deliberation_id.clone(),
-                from_node_ids: vec![root_id.clone()],
-                strategy: Some(ExpandStrategy::Diverse),
-                count: Some(3),
+                node_ids: Vec::new(),
+                critical_claims: Vec::new(),
+                claims: Some(vec![
+                    VerifyClaimInput {
+                        text: "Release requires approval".to_string(),
+                        critical: true,
+                        evidence: Vec::new(),
+                    },
+                    VerifyClaimInput {
+                        text: "Secondary claim".to_string(),
+                        critical: false,
+                        evidence: Vec::new(),
+                    },
+                ]),
+                method: Some(VerificationMethod::Hybrid),
+                policy_override: None,
             })
-            .expect("expand should work");
-        let first = expanded.created_nodes[0].node_id.clone();
-        let second = expanded.created_nodes[1].node_id.clone();
+            .expect("initial verification");
 
-        let session = store
-            .deliberations
-            .get(&started.deliberation_id)
-            .expect("session");
-        assert!(
-            session
-                .children_by_parent
-                .get(&root_id)
+        let updated = store
+            .verify(VerifyThoughtsInput {
+                deliberation_id: started.deliberation_id,
+                node_ids: Vec::new(),
+                critical_claims: Vec::new(),
+                claims: Some(vec![VerifyClaimInput {
+                    text: "  Secondary   claim  ".to_string(),
+                    critical: false,
+                    evidence: vec![qualifying_tier1_evidence("secondary-source")],
+                }]),
+                method: Some(VerificationMethod::Hybrid),
+                policy_override: None,
+            })
+            .expect("verification upsert");
+
+        assert_eq!(updated.verification_matrix.len(), 2);
+        assert!(updated.verification_matrix.iter().any(|entry| {
+            entry.claim == "Release requires approval"
+                && entry.critical
+                && matches!(entry.status, VerificationStatus::DataDeficit)
+        }));
+        assert!(updated.verification_matrix.iter().any(|entry| {
+            entry.claim == "Secondary claim"
+                && !entry.critical
+                && matches!(entry.status, VerificationStatus::Verified)
+        }));
+        assert_eq!(updated.claim_status_summary.verified, 1);
+        assert_eq!(updated.claim_status_summary.data_deficit, 1);
+    }
+
+    #[test]
+    fn verification_upsert_preserves_criticality_and_rechecks_evidence() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("verification-criticality-monotonic".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Keep critical release claims fail closed".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Paranoid),
+            limits: None,
+            verification_policy: None,
+        });
+
+        store
+            .verify(VerifyThoughtsInput {
+                deliberation_id: started.deliberation_id.clone(),
+                node_ids: Vec::new(),
+                critical_claims: Vec::new(),
+                claims: Some(vec![VerifyClaimInput {
+                    text: "Release requires approval".to_string(),
+                    critical: true,
+                    evidence: Vec::new(),
+                }]),
+                method: Some(VerificationMethod::Hybrid),
+                policy_override: None,
+            })
+            .expect("initial critical verification");
+
+        let evidence = (1..=3)
+            .map(|index| EvidenceItem {
+                source: format!("secondary-source-{index}"),
+                tier: EvidenceTier::Tier2,
+                independence_group: format!("secondary-group-{index}"),
+                supports: true,
+                contradictory: false,
+                unambiguous: true,
+            })
+            .collect();
+        let updated = store
+            .verify(VerifyThoughtsInput {
+                deliberation_id: started.deliberation_id.clone(),
+                node_ids: Vec::new(),
+                critical_claims: Vec::new(),
+                claims: Some(vec![VerifyClaimInput {
+                    text: "  Release   requires approval  ".to_string(),
+                    critical: false,
+                    evidence,
+                }]),
+                method: Some(VerificationMethod::Hybrid),
+                policy_override: None,
+            })
+            .expect("same-claim verification upsert");
+
+        assert_eq!(updated.verification_matrix.len(), 1);
+        let entry = &updated.verification_matrix[0];
+        assert!(entry.critical, "criticality must be monotonic");
+        assert!(matches!(entry.status, VerificationStatus::DataDeficit));
+
+        let consensus = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id,
+                method: None,
+                policy_override: None,
+            })
+            .expect("consensus returns a policy-blocked result");
+        assert!(consensus.policy_blocked);
+        assert!(matches!(
+            consensus.route_decision,
+            RouteDecision::GatherMoreEvidence
+        ));
+    }
+
+    #[test]
+    fn conflicting_assumption_criticality_cannot_bypass_governance() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("assumption-criticality-conflict".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Do not downgrade an explicitly critical assumption".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+
+        let recorded = store
+            .record_assumption(RecordAssumptionInput {
+                deliberation_id: started.deliberation_id.clone(),
+                text: "Deployment safety must be established".to_string(),
+                source_node_ids: Vec::new(),
+                confidence: Some(0.99),
+                criticality: Some(AssumptionCriticality::Medium),
+                critical: Some(true),
+                status: Some(AssumptionStatus::Verified),
+                verifiable: Some(true),
+                depends_on: Vec::new(),
+                invalidates: Vec::new(),
+                evidence: Vec::new(),
+                notes: Some("conflicting caller fields".to_string()),
+            })
+            .expect("record conflicting criticality input");
+
+        assert!(matches!(
+            recorded.assumption.criticality,
+            AssumptionCriticality::Critical
+        ));
+        assert!(matches!(
+            recorded.assumption.status,
+            AssumptionStatus::Unresolved
+        ));
+        assert_eq!(recorded.blocking_assumption_ids.len(), 1);
+
+        let checkpoint = store
+            .run_reasoning_checkpoint(RunReasoningCheckpointInput {
+                deliberation_id: started.deliberation_id.clone(),
+                label: Some("criticality conflict gate".to_string()),
+                fail_threshold: Some(0.0),
+            })
+            .expect("checkpoint");
+        assert!(matches!(
+            checkpoint.checkpoint.route_decision,
+            RouteDecision::GatherMoreEvidence
+        ));
+        assert!(!checkpoint.checkpoint.blocking_gaps.is_empty());
+
+        let consensus = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id,
+                method: None,
+                policy_override: None,
+            })
+            .expect("blocked consensus result");
+        assert!(consensus.policy_blocked);
+        assert!(matches!(
+            consensus.route_decision,
+            RouteDecision::GatherMoreEvidence
+        ));
+    }
+
+    #[test]
+    fn consensus_blocks_scaffold_only_frontiers_without_persisting_override() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("scaffold-consensus-gate".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Require agent-authored reasoning before consensus".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+        let root = store
+            .get(&started.deliberation_id)
+            .and_then(|session| session.frontier.first())
+            .cloned()
+            .expect("root");
+        store
+            .expand(ExpandThoughtsInput {
+                deliberation_id: started.deliberation_id.clone(),
+                from_node_ids: vec![root],
+                strategy: Some(ExpandStrategy::Diverse),
+                count: Some(2),
+            })
+            .expect("create scaffold-only frontier");
+
+        let consensus = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id.clone(),
+                method: None,
+                policy_override: Some(verification_policy_from_text("relaxed")),
+            })
+            .expect("consensus returns a blocked result");
+
+        assert!(consensus.policy_blocked);
+        assert!(matches!(
+            consensus.route_decision,
+            RouteDecision::GatherMoreEvidence
+        ));
+        let session = store.get(&started.deliberation_id).expect("session");
+        assert!(
+            session
+                .verification_policy
+                .fail_closed_on_critical_unresolved
+        );
+
+        let wire = serde_json::to_value(consensus_blocked_tool_result(&consensus))
+            .expect("blocked consensus wire result");
+        assert_eq!(wire["isError"], true);
+        assert_eq!(wire["structuredContent"]["code"], "consensus_blocked");
+        let rendered = serde_json::to_string(&wire).expect("wire JSON");
+        assert!(!rendered.contains("final_answer"));
+        assert!(!rendered.contains("Require agent-authored reasoning before consensus"));
+    }
+
+    #[test]
+    fn consensus_requires_an_authored_frontier_and_never_promotes_scaffolds() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("authored-consensus-candidates".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Return only host-authored decision content".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+
+        let root_only = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id.clone(),
+                method: None,
+                policy_override: None,
+            })
+            .expect("root-only consensus returns a blocked route");
+        assert!(root_only.policy_blocked);
+
+        let spoof_error = store
+            .record_reasoning_action(RecordReasoningActionInput {
+                deliberation_id: started.deliberation_id.clone(),
+                tool_name: "rg".to_string(),
+                parameters: Some(json!({"pattern": "release"})),
+                parent_node_id: None,
+                status: Some("planned".to_string()),
+                branch_id: None,
+                tags: Some(vec!["agent-authored".to_string()]),
+                add_to_frontier: Some(true),
+            })
+            .expect_err("agent-authored is a reserved action tag");
+        assert!(spoof_error.contains("agent-authored"));
+        let action_spoof = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id.clone(),
+                method: None,
+                policy_override: None,
+            })
+            .expect("rejected action-tag spoof leaves a blocked root-only route");
+        assert!(action_spoof.policy_blocked);
+
+        let root = store
+            .get(&started.deliberation_id)
+            .and_then(|session| session.frontier.first())
+            .cloned()
+            .expect("root");
+        let expanded = store
+            .expand(ExpandThoughtsInput {
+                deliberation_id: started.deliberation_id.clone(),
+                from_node_ids: vec![root],
+                strategy: Some(ExpandStrategy::Diverse),
+                count: Some(1),
+            })
+            .expect("scaffold expansion");
+        let scaffold_id = expanded.created_nodes[0].node_id.clone();
+        let authored = store
+            .add_thought_node(AddThoughtNodeInput {
+                deliberation_id: started.deliberation_id.clone(),
+                content: "Ship only after the signed release evidence is attached.".to_string(),
+                parent_node_id: Some(scaffold_id),
+                edge_kind: Some(GraphEdgeKind::Refines),
+                edge_note: Some("host-authored replacement".to_string()),
+                tags: None,
+                branch_id: None,
+                add_to_frontier: Some(true),
+            })
+            .expect("agent-authored frontier node");
+
+        let consensus = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id,
+                method: None,
+                policy_override: None,
+            })
+            .expect("authored consensus");
+        assert!(!consensus.policy_blocked);
+        assert!(consensus.final_answer.contains("signed release evidence"));
+        assert!(!consensus.final_answer.contains("SCAFFOLD"));
+        assert!(!consensus.final_answer.contains("Hypothesis:"));
+        assert!(!consensus.final_answer.contains("Return only host-authored"));
+        assert_eq!(consensus.analysis.source_node_ids, vec![authored.node_id]);
+    }
+
+    #[test]
+    fn markdown_table_cells_neutralize_active_markdown() {
+        let escaped = markdown_table_cell(
+            "![pixel](https://attacker.invalid/audit.png) [click](https://attacker.invalid) `code` *bold* _under_",
+        );
+        for active in ["![", "](", "`code`", "*bold*", "_under_"] {
+            assert!(
+                !escaped.contains(active),
+                "active Markdown remained: {active}"
+            );
+        }
+        assert!(escaped.contains("&#33;&#91;pixel&#93;&#40;"));
+        assert!(escaped.contains("&#96;code&#96;"));
+    }
+
+    #[test]
+    fn verified_assumptions_require_qualifying_evidence() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("assumption-evidence-gate".to_string()),
+            mode: Some(ReasoningMode::Got),
+            goal: "Do not self-certify assumptions".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Paranoid),
+            limits: None,
+            verification_policy: None,
+        });
+
+        let recorded = store
+            .record_assumption(RecordAssumptionInput {
+                deliberation_id: started.deliberation_id.clone(),
+                text: "Critical dependency is available".to_string(),
+                source_node_ids: Vec::new(),
+                confidence: Some(0.99),
+                criticality: Some(AssumptionCriticality::Critical),
+                critical: Some(true),
+                status: Some(AssumptionStatus::Verified),
+                verifiable: Some(true),
+                depends_on: Vec::new(),
+                invalidates: Vec::new(),
+                evidence: Vec::new(),
+                notes: None,
+            })
+            .expect("record assumption");
+        assert!(matches!(
+            recorded.assumption.status,
+            AssumptionStatus::Unresolved
+        ));
+        assert_eq!(recorded.blocking_assumption_ids.len(), 1);
+
+        let insufficient = store
+            .set_assumption_status(SetAssumptionStatusInput {
+                deliberation_id: started.deliberation_id.clone(),
+                assumption_id: recorded.assumption.assumption_id.clone(),
+                status: AssumptionStatus::Verified,
+                confidence: None,
+                evidence: Some(vec![EvidenceItem {
+                    source: "secondary-source".to_string(),
+                    tier: EvidenceTier::Tier2,
+                    independence_group: "secondary-source".to_string(),
+                    supports: true,
+                    contradictory: false,
+                    unambiguous: true,
+                }]),
+                notes: None,
+            })
+            .expect("attempt evidence-deficient verification");
+        assert!(matches!(
+            insufficient.assumption.status,
+            AssumptionStatus::Unresolved
+        ));
+        assert_eq!(insufficient.blocking_assumption_ids.len(), 1);
+
+        let verified = store
+            .set_assumption_status(SetAssumptionStatusInput {
+                deliberation_id: started.deliberation_id,
+                assumption_id: recorded.assumption.assumption_id,
+                status: AssumptionStatus::Verified,
+                confidence: Some(0.95),
+                evidence: Some(vec![qualifying_tier1_evidence("official-source")]),
+                notes: None,
+            })
+            .expect("evidence-backed verification");
+        assert!(matches!(
+            verified.assumption.status,
+            AssumptionStatus::Verified
+        ));
+        assert!(verified.blocking_assumption_ids.is_empty());
+    }
+
+    #[test]
+    fn reasoning_observation_requires_action_and_completes_it_atomically() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("react-observation-integrity".to_string()),
+            mode: Some(ReasoningMode::Got),
+            goal: "Keep ReAct state consistent".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+        let root = store
+            .get(&started.deliberation_id)
+            .and_then(|session| session.frontier.first())
+            .cloned()
+            .expect("root");
+
+        let wrong_target = store.record_reasoning_observation(RecordReasoningObservationInput {
+            deliberation_id: started.deliberation_id.clone(),
+            action_node_id: root,
+            result: Some(json!({"ok": false})),
+        });
+        assert!(
+            wrong_target
+                .expect_err("thought nodes cannot own observations")
+                .contains("Action")
+        );
+
+        let action = store
+            .record_reasoning_action(RecordReasoningActionInput {
+                deliberation_id: started.deliberation_id.clone(),
+                tool_name: "unit-test".to_string(),
+                parameters: Some(json!({"input": "value"})),
+                parent_node_id: None,
+                status: Some("planned".to_string()),
+                branch_id: None,
+                tags: None,
+                add_to_frontier: Some(false),
+            })
+            .expect("record action");
+
+        let node_count = store
+            .get(&started.deliberation_id)
+            .map(|session| session.nodes.len())
+            .expect("session");
+        store
+            .deliberations
+            .get_mut(&started.deliberation_id)
+            .expect("session")
+            .limits
+            .max_nodes = node_count as i64;
+        let capacity_failure =
+            store.record_reasoning_observation(RecordReasoningObservationInput {
+                deliberation_id: started.deliberation_id.clone(),
+                action_node_id: action.action_node_id.clone(),
+                result: Some(json!({"ok": true})),
+            });
+        assert!(
+            capacity_failure
+                .expect_err("capacity failure")
+                .contains("max_nodes")
+        );
+        let action_after_failure =
+            &store.get(&started.deliberation_id).expect("session").nodes[&action.action_node_id];
+        assert!(matches!(
+            action_after_failure.node_type,
+            ThoughtNodeType::Action {
+                status: ActionStatus::Planned,
+                ..
+            }
+        ));
+        assert!(action_after_failure.content.contains("status=Planned"));
+        assert_eq!(
+            store
+                .get(&started.deliberation_id)
+                .expect("session")
+                .nodes
+                .len(),
+            node_count
+        );
+
+        store
+            .deliberations
+            .get_mut(&started.deliberation_id)
+            .expect("session")
+            .limits
+            .max_nodes += 1;
+        let observation = store
+            .record_reasoning_observation(RecordReasoningObservationInput {
+                deliberation_id: started.deliberation_id.clone(),
+                action_node_id: action.action_node_id.clone(),
+                result: Some(json!({"ok": true})),
+            })
+            .expect("record observation");
+        let session = store.get(&started.deliberation_id).expect("session");
+        assert!(session.nodes.contains_key(&observation.observation_node_id));
+        assert!(matches!(
+            session.nodes[&action.action_node_id].node_type,
+            ThoughtNodeType::Action {
+                status: ActionStatus::Complete,
+                ..
+            }
+        ));
+        assert!(
+            session.nodes[&action.action_node_id]
+                .content
+                .contains("status=Complete")
+        );
+    }
+
+    #[test]
+    fn unknown_pipeline_stage_maps_to_invalid_input() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("unknown-pipeline-stage".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Reject misspelled governance stages".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+
+        let error = store
+            .run_pipeline(RunReasonKitPipelineInput {
+                deliberation_id: started.deliberation_id,
+                stages: Some(vec!["triangulaton".to_string()]),
+                profile: None,
+                policy_override: None,
+            })
+            .expect_err("unknown stage must fail closed");
+        let envelope = ErrorEnvelope::from_store_error(error);
+        assert_eq!(envelope.code, "invalid_input");
+    }
+
+    #[test]
+    fn rejected_pipeline_call_leaves_session_unchanged() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("rejected-pipeline-transaction".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Reject invalid pipeline requests atomically".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Paranoid),
+            limits: None,
+            verification_policy: None,
+        });
+        let before = serde_json::to_value(
+            store
+                .get(&started.deliberation_id)
+                .expect("deliberation session"),
+        )
+        .expect("serialize session before rejected call");
+
+        let error = store
+            .run_pipeline(RunReasonKitPipelineInput {
+                deliberation_id: started.deliberation_id.clone(),
+                stages: Some(vec!["unknown-stage".to_string()]),
+                profile: None,
+                policy_override: Some(verification_policy_from_text("relaxed")),
+            })
+            .expect_err("unknown stage must fail closed");
+        assert!(error.contains("invalid pipeline stage"));
+
+        let after = serde_json::to_value(
+            store
+                .get(&started.deliberation_id)
+                .expect("deliberation session"),
+        )
+        .expect("serialize session after rejected call");
+        assert_eq!(after, before, "failed calls must not retain partial state");
+    }
+
+    #[test]
+    fn pipeline_policy_override_is_call_scoped() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("pipeline-call-scoped-policy".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Keep session verification policy explicit".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Paranoid),
+            limits: None,
+            verification_policy: None,
+        });
+        let policy_before = serde_json::to_value(
+            &store
+                .get(&started.deliberation_id)
+                .expect("deliberation session")
+                .verification_policy,
+        )
+        .expect("serialize policy before pipeline call");
+
+        store
+            .run_pipeline(RunReasonKitPipelineInput {
+                deliberation_id: started.deliberation_id.clone(),
+                stages: Some(vec!["causal-modeling".to_string()]),
+                profile: None,
+                policy_override: Some(verification_policy_from_text("relaxed")),
+            })
+            .expect("known pipeline stage");
+
+        let policy_after = serde_json::to_value(
+            &store
+                .get(&started.deliberation_id)
+                .expect("deliberation session")
+                .verification_policy,
+        )
+        .expect("serialize policy after pipeline call");
+        assert_eq!(
+            policy_after, policy_before,
+            "pipeline policy overrides must not persist on the session"
+        );
+    }
+
+    #[test]
+    fn audit_export_uses_one_id_for_envelope_and_payload() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("audit-id-integrity".to_string()),
+            mode: Some(ReasoningMode::Reasonkit),
+            goal: "Export a correlatable audit".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+
+        let audit = store
+            .export_audit(ExportReasoningAuditInput {
+                deliberation_id: started.deliberation_id,
+                format: AuditFormat::Json,
+                include_raw_thoughts: Some(false),
+            })
+            .expect("audit export");
+        assert_eq!(
+            audit.payload["audit_id"].as_str(),
+            Some(audit.audit_id.as_str())
+        );
+    }
+
+    #[test]
+    fn audit_export_defaults_to_canonical_json_artifact() {
+        let (store, deliberation_id) = seeded_audit_store();
+        let input: ExportReasoningAuditInput = serde_json::from_value(json!({
+            "deliberation_id": deliberation_id,
+            "include_raw_thoughts": false
+        }))
+        .expect("legacy audit input without format");
+
+        let result = serde_json::to_value(store.export_audit(input).expect("JSON audit"))
+            .expect("serialize audit result");
+        assert_eq!(result["format"], json!("json"));
+        assert_eq!(result["content_type"], json!("application/json"));
+        let artifact = result["artifact"].as_str().expect("JSON artifact");
+        let artifact_payload: Value = serde_json::from_str(artifact).expect("valid JSON artifact");
+        assert_eq!(artifact_payload, result["payload"]);
+        assert_eq!(result["audit_id"], result["payload"]["audit_id"]);
+        assert!(result["payload"].get("nodes").is_none());
+        assert!(result["payload"].get("format").is_none());
+        assert!(result["payload"].get("artifact").is_none());
+    }
+
+    #[test]
+    fn markdown_audit_is_complete_sorted_escaped_and_redacted_by_default() {
+        let (store, deliberation_id) = seeded_audit_store();
+        let input: ExportReasoningAuditInput = serde_json::from_value(json!({
+            "deliberation_id": deliberation_id,
+            "format": "markdown"
+        }))
+        .expect("Markdown audit input");
+
+        let result = serde_json::to_value(store.export_audit(input).expect("Markdown audit"))
+            .expect("serialize audit result");
+        assert_eq!(result["format"], json!("markdown"));
+        assert_eq!(
+            result["content_type"],
+            json!("text/markdown; charset=utf-8")
+        );
+        let artifact = result["artifact"].as_str().expect("Markdown artifact");
+        for required in [
+            "# Reasoning Audit",
+            "audit-session",
+            "reasonkit",
+            "GATHER&#95;MORE&#95;EVIDENCE",
+            "## Verification matrix",
+            "verified",
+            "data&#95;deficit",
+            "z-source",
+            "## Assumptions",
+            "unresolved",
+            "critical",
+            "## Quality checkpoints",
+            "checkpoint blocker",
+            "## Pipeline gates",
+            "pipeline blocker",
+            "HEURISTIC OUTPUT",
+            "reasonkit-think/heuristic",
+            "host&#95;agent",
+            "Raw node content omitted",
+        ] {
+            assert!(
+                artifact.contains(required),
+                "missing `{required}`\n{artifact}"
+            );
+        }
+        assert!(artifact.contains("Alpha \\| claim<br>next &lt;tag&gt;"));
+        assert!(artifact.contains("a\\|source"));
+        assert!(artifact.contains("a \\| checkpoint blocker"));
+        assert!(artifact.contains("a \\| pipeline blocker"));
+        assert!(!artifact.contains("RAW_ALPHA_NODE_CONTENT"));
+        assert!(!artifact.contains("RAW_ZETA_NODE_CONTENT"));
+        assert!(
+            artifact.find("assumption-a").expect("assumption-a")
+                < artifact.find("assumption-z").expect("assumption-z")
+        );
+        assert!(
+            artifact.find("checkpoint-a").expect("checkpoint-a")
+                < artifact.find("checkpoint-z").expect("checkpoint-z")
+        );
+        assert!(
+            artifact.find("a-stage").expect("a-stage") < artifact.find("z-stage").expect("z-stage")
+        );
+        assert!(result["payload"].get("nodes").is_none());
+    }
+
+    #[test]
+    fn markdown_audit_includes_sorted_raw_nodes_only_when_requested() {
+        let (store, deliberation_id) = seeded_audit_store();
+        let input: ExportReasoningAuditInput = serde_json::from_value(json!({
+            "deliberation_id": deliberation_id,
+            "format": "markdown",
+            "include_raw_thoughts": true
+        }))
+        .expect("raw Markdown audit input");
+
+        let result = serde_json::to_value(store.export_audit(input).expect("raw Markdown audit"))
+            .expect("serialize audit result");
+        let artifact = result["artifact"].as_str().expect("Markdown artifact");
+        assert!(result["payload"]["nodes"].is_object());
+        assert!(
+            artifact.contains(
+                "RAW&#95;ALPHA&#95;NODE&#95;CONTENT \\| line<br>&lt;script&gt;alert&#40;1&#41;&lt;/script&gt;"
+            )
+        );
+        assert!(artifact.contains("RAW&#95;ZETA&#95;NODE&#95;CONTENT"));
+        assert!(
+            artifact.find("node-a").expect("node-a") < artifact.find("node-z").expect("node-z")
+        );
+    }
+
+    #[test]
+    fn audit_format_rejects_values_outside_json_and_markdown() {
+        let parsed = serde_json::from_value::<ExportReasoningAuditInput>(json!({
+            "deliberation_id": "delib-test",
+            "format": "html"
+        }));
+        assert!(
+            parsed.is_err(),
+            "unsupported formats must fail input parsing"
+        );
+    }
+
+    #[test]
+    fn graph_indexes_track_lineage_and_strict_cycle_rejections() {
+        let mut store = DeliberationStore::new(64);
+        let started = store.start(StartDeliberationInput {
+            session_id: Some("graph-indexes".to_string()),
+            mode: Some(ReasoningMode::Got),
+            goal: "Maintain graph indexes".to_string(),
+            constraints: None,
+            profile: Some(ReasoningProfile::Balanced),
+            limits: None,
+            verification_policy: None,
+        });
+        let root_id = store
+            .deliberations
+            .get(&started.deliberation_id)
+            .and_then(|session| session.frontier.first())
+            .cloned()
+            .expect("root frontier");
+
+        let expanded = store
+            .expand(ExpandThoughtsInput {
+                deliberation_id: started.deliberation_id.clone(),
+                from_node_ids: vec![root_id.clone()],
+                strategy: Some(ExpandStrategy::Diverse),
+                count: Some(3),
+            })
+            .expect("expand should work");
+        let first = expanded.created_nodes[0].node_id.clone();
+        let second = expanded.created_nodes[1].node_id.clone();
+
+        let session = store
+            .deliberations
+            .get(&started.deliberation_id)
+            .expect("session");
+        assert!(
+            session
+                .children_by_parent
+                .get(&root_id)
                 .is_some_and(|children| children.contains(&first))
         );
         assert!(
@@ -15728,6 +17461,7 @@ mod architecture_contract_tests {
                 "reasoning://session/{}/skeleton-phases",
                 started.deliberation_id
             ),
+            ToolPack::Full,
         )
         .expect("phases resource");
         assert_eq!(phases["phase_count"], json!(1));
@@ -15737,6 +17471,7 @@ mod architecture_contract_tests {
                 "reasoning://session/{}/skeleton-phase/{}",
                 started.deliberation_id, phase_id
             ),
+            ToolPack::Full,
         )
         .expect("phase resource");
         assert_eq!(phase["phase"]["state"], json!("MERGED"));
@@ -15744,6 +17479,7 @@ mod architecture_contract_tests {
         let audit = store
             .export_audit(ExportReasoningAuditInput {
                 deliberation_id: started.deliberation_id,
+                format: AuditFormat::Json,
                 include_raw_thoughts: Some(false),
             })
             .expect("audit");
@@ -15838,18 +17574,58 @@ mod architecture_contract_tests {
             RouteDecision::GatherMoreEvidence
         ));
 
+        let rejected = store
+            .set_assumption_status(SetAssumptionStatusInput {
+                deliberation_id: started.deliberation_id.clone(),
+                assumption_id: assumption.assumption.assumption_id.clone(),
+                status: AssumptionStatus::Verified,
+                confidence: Some(0.95),
+                evidence: None,
+                notes: Some("unsupported verification attempt".to_string()),
+            })
+            .expect("unsupported verification remains unresolved");
+        assert!(matches!(
+            rejected.assumption.status,
+            AssumptionStatus::Unresolved
+        ));
+        assert_eq!(rejected.blocking_assumption_ids.len(), 1);
+
+        let still_blocked = store
+            .consensus(ConsensusAnswerInput {
+                deliberation_id: started.deliberation_id.clone(),
+                method: None,
+                policy_override: None,
+            })
+            .expect("consensus remains policy-blocked");
+        assert!(still_blocked.policy_blocked);
+
         let updated = store
             .set_assumption_status(SetAssumptionStatusInput {
                 deliberation_id: started.deliberation_id.clone(),
                 assumption_id: assumption.assumption.assumption_id,
                 status: AssumptionStatus::Verified,
                 confidence: Some(0.95),
-                evidence: None,
-                notes: Some("verified in test".to_string()),
+                evidence: Some(vec![qualifying_tier1_evidence("official-dependency-check")]),
+                notes: Some("verified with qualifying evidence".to_string()),
             })
-            .expect("resolve assumption");
+            .expect("resolve assumption with evidence");
         assert!(updated.blocking_assumption_ids.is_empty());
 
+        store
+            .add_thought_node(AddThoughtNodeInput {
+                deliberation_id: started.deliberation_id.clone(),
+                content:
+                    "Proceed because the official dependency check resolved the critical premise."
+                        .to_string(),
+                parent_node_id: Some(root),
+                edge_kind: Some(GraphEdgeKind::Supports),
+                edge_note: Some("evidence-backed host conclusion".to_string()),
+                tags: None,
+                branch_id: None,
+                add_to_frontier: Some(true),
+            })
+            .expect("add evidence-backed host conclusion");
+
         let unblocked = store
             .consensus(ConsensusAnswerInput {
                 deliberation_id: started.deliberation_id,
@@ -16988,14 +18764,14 @@ mod thinking_mode_contract_tests {
     fn native_resource_catalog_includes_modes_and_templates() {
         let resources = native_reasoning_resources();
         assert!(resources.iter().any(|resource| {
-            resource.raw.uri == "reasoning://thinking-modes"
-                && resource.raw.mime_type.as_deref() == Some("application/json")
+            resource.uri == "reasoning://thinking-modes"
+                && resource.mime_type.as_deref() == Some("application/json")
         }));
         let templates = native_reasoning_resource_templates();
         assert!(
             templates
                 .iter()
-                .any(|template| template.raw.uri_template == "reasoning://session/{id}/graph")
+                .any(|template| template.uri_template == "reasoning://session/{id}/graph")
         );
     }
 }
@@ -17522,8 +19298,371 @@ mod add_thought_node_tests {
     }
 }
 
+#[cfg(test)]
+mod tool_pack_contract_tests {
+    use super::*;
+
+    const CORE: &[&str] = &[
+        "add_thought_node",
+        "consensus_answer",
+        "export_reasoning_audit",
+        "get_reasoning_coaching",
+        "reasoning_intent_router",
+        "record_assumption",
+        "run_reasoning_checkpoint",
+        "run_reasonkit_pipeline",
+        "run_thinking_mode",
+        "set_assumption_status",
+        "set_verification_policy",
+        "start_deliberation",
+        "verify_thoughts",
+    ];
+
+    const STANDARD: &[&str] = &[
+        "add_thought_node",
+        "apply_reasoning_pattern",
+        "consensus_answer",
+        "converge_reasoning",
+        "distill_thought_cluster",
+        "evaluate_reasoning_quality",
+        "expand_thoughts",
+        "export_memory_snapshot",
+        "export_reasoning_audit",
+        "get_reasoning_coaching",
+        "get_thinking_history",
+        "link_thoughts",
+        "list_reasoning_resources",
+        "merge_thought_branches",
+        "plan_tool_sequence",
+        "prune_thoughts",
+        "read_reasoning_resource",
+        "reasoning_intent_router",
+        "record_assumption",
+        "refine_thoughts",
+        "replay_reasoning_session",
+        "run_reasoning_checkpoint",
+        "run_reasonkit_pipeline",
+        "run_skeleton_elaboration",
+        "run_thinking_mode",
+        "score_thoughts",
+        "set_assumption_status",
+        "set_reasoning_aliases",
+        "set_scoring_rubric",
+        "set_verification_policy",
+        "start_deliberation",
+        "start_skeleton_of_thought",
+        "verify_thoughts",
+    ];
+
+    const FULL: &[&str] = &[
+        "add_thought_node",
+        "apply_algorithm_template",
+        "apply_reasoning_pattern",
+        "clear_thinking_history",
+        "consensus_answer",
+        "converge_reasoning",
+        "distill_thought_cluster",
+        "evaluate_reasoning_quality",
+        "expand_thoughts",
+        "export_memory_snapshot",
+        "export_reasoning_audit",
+        "get_reasoning_coaching",
+        "get_thinking_history",
+        "link_thoughts",
+        "list_reasoning_resources",
+        "mcts_select_path",
+        "merge_thought_branches",
+        "plan_tool_sequence",
+        "prune_thoughts",
+        "query_failure_memory",
+        "read_reasoning_resource",
+        "reasoning_autopilot",
+        "reasoning_intent_router",
+        "record_assumption",
+        "record_reasoning_action",
+        "record_reasoning_observation",
+        "refine_thoughts",
+        "replay_reasoning_session",
+        "run_reasoning_checkpoint",
+        "run_reasonkit_pipeline",
+        "run_skeleton_elaboration",
+        "run_thinking_mode",
+        "score_thoughts",
+        "sequentialthinking_tools",
+        "set_assumption_status",
+        "set_reasoning_aliases",
+        "set_scoring_rubric",
+        "set_verification_policy",
+        "start_deliberation",
+        "start_skeleton_of_thought",
+        "think_query",
+        "verify_thoughts",
+        "web_research_enqueue",
+        "web_research_fetch",
+        "web_research_task_status",
+        "web_research_triangulate",
+        "web_research_verify",
+    ];
+
+    fn router_names(server: &ThinkServer) -> Vec<String> {
+        server
+            .tool_router
+            .list_all()
+            .into_iter()
+            .map(|tool| tool.name.into_owned())
+            .collect()
+    }
+
+    #[test]
+    fn tool_pack_members_match_the_adoption_contract() {
+        assert_eq!(ToolPack::Core.members(), CORE);
+        assert_eq!(ToolPack::Standard.members(), STANDARD);
+        assert_eq!(ToolPack::Full.members(), FULL);
+        assert_eq!(ThinkServer::tool_router().list_all().len(), FULL.len());
+        assert!(CORE.iter().all(|tool| STANDARD.contains(tool)));
+        assert!(STANDARD.iter().all(|tool| FULL.contains(tool)));
+    }
+
+    #[test]
+    fn tool_pack_resolution_defaults_and_falls_back_to_core() {
+        assert_eq!(ToolPack::resolve(None), (ToolPack::Core, None));
+        assert_eq!(ToolPack::resolve(Some("")), (ToolPack::Core, None));
+        assert_eq!(ToolPack::resolve(Some("core")), (ToolPack::Core, None));
+        assert_eq!(
+            ToolPack::resolve(Some("standard")),
+            (ToolPack::Standard, None)
+        );
+        assert_eq!(ToolPack::resolve(Some("full")), (ToolPack::Full, None));
+        let (pack, warning) = ToolPack::resolve(Some("everything"));
+        assert_eq!(pack, ToolPack::Core);
+        assert_eq!(
+            warning,
+            Some(
+                "invalid REASONKIT_TOOL_PACK; expected core, standard, or full; using safe core fallback"
+            )
+        );
+    }
+
+    #[test]
+    fn stored_router_advertises_exact_tools_for_each_pack() {
+        let core = ThinkServer::new_with_tool_pack(20, ToolPack::Core);
+        let standard = ThinkServer::new_with_tool_pack(20, ToolPack::Standard);
+        let full = ThinkServer::new_with_tool_pack(20, ToolPack::Full);
+        assert_eq!(router_names(&core), CORE);
+        assert_eq!(router_names(&standard), STANDARD);
+        assert_eq!(router_names(&full), FULL);
+    }
+
+    #[test]
+    fn tool_pack_resource_reports_active_and_exact_memberships() {
+        let resource = tool_pack_resource(ToolPack::Standard);
+        assert_eq!(resource["default"], "core");
+        assert_eq!(resource["active"], "standard");
+        assert_eq!(resource["active_tool_count"], 33);
+        assert_eq!(resource["restart_required"], true);
+        assert_eq!(resource["packs"]["core"]["tools"], json!(CORE));
+        assert_eq!(resource["packs"]["standard"]["tools"], json!(STANDARD));
+        assert_eq!(resource["packs"]["full"]["tools"], json!(FULL));
+    }
+
+    #[test]
+    fn core_guidance_marks_hidden_tools_instead_of_silently_recommending_them() {
+        let server = ThinkServer::new_with_tool_pack(20, ToolPack::Core);
+        let info = server.get_info();
+        let instructions = info.instructions.expect("server instructions");
+        assert!(instructions.contains("active tool pack: core (13 tools)"));
+        assert!(instructions.contains("REASONKIT_TOOL_PACK=full"));
+
+        let prompt = server.sequential_thinking_guidance(Parameters(GuidancePromptInput {
+            problem: Some("small task".to_string()),
+        }));
+        let prompt_json = serde_json::to_value(prompt).expect("prompt json");
+        let prompt_text = prompt_json["messages"][0]["content"]["text"]
+            .as_str()
+            .expect("prompt text");
+        assert!(prompt_text.contains("sequentialthinking_tools requires the full tool pack"));
+        assert!(prompt_text.contains("run_thinking_mode"));
+
+        let modes = thinking_modes_resource(&ReasoningAliases::default(), ToolPack::Core);
+        assert_eq!(modes["active_tool_pack"], "core");
+        for mode in modes["modes"].as_array().expect("mode array") {
+            for tool in mode["primary_tools"].as_array().expect("primary tools") {
+                assert!(CORE.contains(&tool.as_str().expect("tool name")));
+            }
+        }
+        assert_eq!(modes["compatibility_default_tool"], Value::Null);
+    }
+
+    #[test]
+    fn json_tool_results_are_dual_format_and_fail_closed() {
+        let payload = json!({"answer": 42, "status": "ok"});
+        let success = serde_json::to_value(json_tool_success(&payload)).expect("success wire json");
+        assert_eq!(success["structuredContent"], payload);
+        assert_eq!(success["isError"], false);
+        let success_text = success["content"][0]["text"]
+            .as_str()
+            .expect("success text");
+        assert_eq!(
+            serde_json::from_str::<Value>(success_text).unwrap(),
+            payload
+        );
+
+        let failure = serde_json::to_value(json_tool_error(&payload)).expect("error wire json");
+        assert_eq!(failure["structuredContent"], payload);
+        assert_eq!(failure["isError"], true);
+        let failure_text = failure["content"][0]["text"].as_str().expect("error text");
+        assert_eq!(
+            serde_json::from_str::<Value>(failure_text).unwrap(),
+            payload
+        );
+
+        struct BrokenSerialize;
+        impl Serialize for BrokenSerialize {
+            fn serialize<S>(&self, _serializer: S) -> Result<S::Ok, S::Error>
+            where
+                S: serde::Serializer,
+            {
+                Err(serde::ser::Error::custom(
+                    "intentional serialization failure",
+                ))
+            }
+        }
+
+        let broken =
+            serde_json::to_value(json_tool_success(&BrokenSerialize)).expect("fallback wire json");
+        assert_eq!(broken["isError"], true);
+        assert_eq!(broken["structuredContent"]["code"], "serialization_error");
+        let broken_text = broken["content"][0]["text"]
+            .as_str()
+            .expect("fallback text");
+        assert_eq!(
+            serde_json::from_str::<Value>(broken_text).unwrap(),
+            broken["structuredContent"]
+        );
+    }
+}
+
+#[cfg(test)]
+mod cli_contract_tests {
+    use super::*;
+
+    #[test]
+    fn cli_actions_preserve_stdio_default_and_reject_unknown_arguments() {
+        assert_eq!(parse_cli_action(&[]).unwrap(), CliAction::ServeStdio);
+        assert_eq!(
+            parse_cli_action(&["--version".to_string()]).unwrap(),
+            CliAction::Version
+        );
+        assert_eq!(
+            parse_cli_action(&["--help".to_string()]).unwrap(),
+            CliAction::Help
+        );
+        assert_eq!(
+            parse_cli_action(&["doctor".to_string(), "--json".to_string()]).unwrap(),
+            CliAction::DoctorJson
+        );
+        assert!(parse_cli_action(&["--unknown".to_string()]).is_err());
+    }
+
+    #[test]
+    fn doctor_report_is_pack_aware_and_exposes_no_secret_values() {
+        let report = doctor_report(Some("standard"), BackendMode::resolve(None));
+        assert_eq!(report["status"], "ok");
+        assert_eq!(report["version"], env!("CARGO_PKG_VERSION"));
+        assert_eq!(report["transport"], "stdio");
+        assert_eq!(report["active_tool_pack"], "standard");
+        assert_eq!(report["tool_count"], 33);
+        assert_eq!(report["semantic_engine"], "host_agent");
+        assert_eq!(report["backend_mode"], "agent");
+        assert_eq!(report["external_provider_bound"], false);
+        assert_eq!(report["retrieval_provider_bound"], false);
+        let rendered = serde_json::to_string(&report).expect("doctor JSON");
+        for forbidden in ["TOKEN", "API_KEY", "BEARER", "password"] {
+            assert!(!rendered.contains(forbidden));
+        }
+
+        let fallback = doctor_report(Some("misspelled"), BackendMode::resolve(None));
+        assert_eq!(fallback["status"], "warning");
+        assert_eq!(fallback["active_tool_pack"], "core");
+        assert_eq!(fallback["tool_count"], 13);
+        assert!(
+            fallback["warnings"][0]
+                .as_str()
+                .expect("warning")
+                .contains("invalid REASONKIT_TOOL_PACK")
+        );
+    }
+
+    #[test]
+    fn doctor_report_accepts_unset_and_agent_backend_modes() {
+        for raw_backend_mode in [None, Some("agent")] {
+            let report = doctor_report(None, BackendMode::resolve(raw_backend_mode));
+            assert_eq!(report["status"], "ok");
+            assert_eq!(report["backend_mode"], "agent");
+            assert_eq!(report["external_provider_bound"], false);
+            assert_eq!(report["warnings"], json!([]));
+        }
+    }
+
+    #[test]
+    fn doctor_report_warns_and_falls_back_for_invalid_backend_mode_without_echoing_it() {
+        let invalid = "external-with-secret-7f3a9c";
+        let report = doctor_report(None, BackendMode::resolve(Some(invalid)));
+
+        assert_eq!(report["status"], "warning");
+        assert_eq!(report["backend_mode"], "agent");
+        assert_eq!(report["external_provider_bound"], false);
+        assert!(
+            report["warnings"][0]
+                .as_str()
+                .expect("warning")
+                .contains("invalid BACKEND_MODE")
+        );
+        assert!(!serde_json::to_string(&report).unwrap().contains(invalid));
+    }
+
+    #[test]
+    fn doctor_report_warns_and_falls_back_when_external_provider_is_unbound() {
+        let report = doctor_report(None, BackendMode::resolve(Some("external")));
+
+        assert_eq!(report["status"], "warning");
+        assert_eq!(report["backend_mode"], "agent");
+        assert_eq!(report["external_provider_bound"], false);
+        assert!(
+            report["warnings"][0]
+                .as_str()
+                .expect("warning")
+                .contains("no external provider is bound")
+        );
+    }
+}
+
 #[tokio::main]
 async fn main() -> anyhow::Result<()> {
+    let args = std::env::args().skip(1).collect::<Vec<_>>();
+    match parse_cli_action(&args).map_err(anyhow::Error::msg)? {
+        CliAction::Version => {
+            println!("reasonkit-think-mcp {}", env!("CARGO_PKG_VERSION"));
+            return Ok(());
+        }
+        CliAction::Help => {
+            println!("{CLI_HELP}");
+            return Ok(());
+        }
+        CliAction::DoctorJson => {
+            let configured_pack = std::env::var("REASONKIT_TOOL_PACK").ok();
+            println!(
+                "{}",
+                serde_json::to_string_pretty(&doctor_report(
+                    configured_pack.as_deref(),
+                    BackendMode::resolve_env()
+                ))?
+            );
+            return Ok(());
+        }
+        CliAction::ServeStdio => {}
+    }
+
     tracing_subscriber::fmt()
         .with_env_filter(
             tracing_subscriber::EnvFilter::try_from_default_env()
@@ -17553,47 +19692,48 @@ async fn main() -> anyhow::Result<()> {
 #[cfg(test)]
 mod reasoning_loop_integration_tests {
     use super::*;
-    use reasonkit_mem::storage::hot::HotMemoryEntry;
     use tokio::runtime::Runtime;
 
     #[test]
-    fn test_reasoning_loop_with_think_query() {
-        let rt = Runtime::new().unwrap();
-        rt.block_on(async {
-            // Initialize ThinkServer
-            let server = ThinkServer::new(100);
-
-            // 1. Add some entries to the hot cache directly (simulating prior thoughts being encoded)
-            let cache = &server.hot_memory;
-            let entry = HotMemoryEntry::new(
-                uuid::Uuid::new_v4(),
-                "Hot cache is blazing fast".to_string(),
-                vec![0.5; 1536],
-                serde_json::json!({"test": true}),
-            );
-            cache.put(entry.clone()).await.expect("hot cache put");
-
-            // 2. Test the think_query fast-path API
-            let query_input = ThinkQueryInput {
+    fn think_query_fails_closed_without_a_bound_provider() {
+        let runtime = Runtime::new().expect("tokio runtime");
+        runtime.block_on(async {
+            let server = ThinkServer::new_with_tool_pack(100, ToolPack::Full);
+            let query_res = server.think_query(Parameters(ThinkQueryInput {
                 query: "fast cache".to_string(),
                 query_embedding: Some(vec![0.5; 1536]),
                 limit: Some(5),
                 min_score: Some(0.1),
-            };
-
-            let query_res = server.think_query(Parameters(query_input));
-            assert_ne!(query_res.is_error, Some(true), "think_query should succeed");
+            }));
 
-            let value = serde_json::to_value(&query_res).expect("result json");
+            assert_eq!(query_res.is_error, Some(true));
+            let value = serde_json::to_value(query_res).expect("result json");
+            assert_eq!(
+                value["structuredContent"]["code"],
+                "experimental_memory_unavailable"
+            );
+            assert_eq!(value["structuredContent"]["retryable"], false);
+            assert!(
+                value["structuredContent"]["error"]
+                    .as_str()
+                    .expect("error message")
+                    .contains("no retrieval was performed")
+            );
             let text = value["content"][0]["text"].as_str().expect("text payload");
-            let json_val = serde_json::from_str::<serde_json::Value>(text).unwrap();
-            let results = json_val.get("results").unwrap().as_array().unwrap();
-            assert!(!results.is_empty(), "Expected results from think_query");
-            let first_result = results[0].as_object().unwrap();
             assert_eq!(
-                first_result.get("content").unwrap().as_str().unwrap(),
-                "Hot cache is blazing fast"
+                serde_json::from_str::<Value>(text).expect("text JSON"),
+                value["structuredContent"]
             );
+
+            let description = server
+                .tool_router
+                .list_all()
+                .into_iter()
+                .find(|tool| tool.name == "think_query")
+                .expect("think_query route")
+                .description
+                .expect("think_query description");
+            assert!(description.contains("unavailable"));
         });
     }
 }
diff --git a/tests/test_ci_contract.py b/tests/test_ci_contract.py
new file mode 100644
index 0000000..9986d81
--- /dev/null
+++ b/tests/test_ci_contract.py
@@ -0,0 +1,81 @@
+from pathlib import Path
+import unittest
+
+
+class CiContractTests(unittest.TestCase):
+    def test_justfile_exposes_unit_eval_and_audit_gates(self) -> None:
+        source = Path("justfile").read_text(encoding="utf-8")
+        self.assertIn("unit:", source)
+        self.assertIn("eval:", source)
+        self.assertIn("audit:", source)
+        self.assertIn("python3 -m unittest discover", source)
+        self.assertIn("evals/run_contract_evals.py", source)
+        self.assertIn("cargo audit --deny warnings", source)
+
+    def test_justfile_exposes_config_only_dist_validation(self) -> None:
+        source = Path("justfile").read_text(encoding="utf-8")
+        self.assertIn("dist-check:", source)
+        self.assertIn(
+            "dist manifest --artifacts=local --output-format=json --no-local-paths",
+            source,
+        )
+
+    def test_justfile_exposes_package_registry_and_dist_release_gates(self) -> None:
+        source = Path("justfile").read_text(encoding="utf-8")
+        self.assertIn("package-check:", source)
+        self.assertIn(
+            'CARGO_TARGET_DIR="{{ justfile_directory() }}/target/package-check" cargo package --locked',
+            source,
+        )
+        self.assertIn("registry-check:", source)
+        self.assertIn("uvx --from check-jsonschema==0.38.0", source)
+        self.assertIn(
+            "ci: check eval audit package-check registry-check dist-check",
+            source,
+        )
+
+    def test_ci_runs_explicit_rust_python_and_runtime_contract_gates(self) -> None:
+        source = Path(".github/workflows/ci.yml").read_text(encoding="utf-8")
+        self.assertIn("cargo test", source)
+        self.assertIn("Contract unit tests", source)
+        self.assertIn("MCP contract evaluations", source)
+        self.assertIn("run: cargo audit --deny warnings", source)
+
+    def test_ci_enforces_pinned_package_registry_and_dist_checks(self) -> None:
+        source = Path(".github/workflows/ci.yml").read_text(encoding="utf-8")
+        self.assertIn(
+            "run: CARGO_TARGET_DIR=target/package-check cargo package --locked",
+            source,
+        )
+        self.assertIn(
+            "uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9",
+            source,
+        )
+        self.assertIn('version: "0.12.1"', source)
+        self.assertIn("uvx --from check-jsonschema==0.38.0", source)
+        self.assertIn(
+            "run: cargo install cargo-dist --version 0.32.0 --locked",
+            source,
+        )
+        self.assertIn(
+            "dist manifest --artifacts=local --output-format=json --no-local-paths",
+            source,
+        )
+
+    def test_audit_warning_cleanup_is_tracked(self) -> None:
+        path = Path("docs/dependency-audit.md")
+        if not path.is_file():
+            self.fail("docs/dependency-audit.md must track the warning-clean boundary")
+        source = path.read_text(encoding="utf-8")
+        self.assertIn("reasonkit-mem", source)
+        self.assertIn("experimental_memory_unavailable", source)
+        self.assertIn("cargo audit --deny warnings", source)
+
+    def test_contributor_audit_copy_matches_the_zero_warning_gate(self) -> None:
+        source = Path("CONTRIBUTING.md").read_text(encoding="utf-8")
+        self.assertNotIn("known warning debt", source)
+        self.assertIn("zero-warning baseline", source)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/test_client_compat_check.py b/tests/test_client_compat_check.py
new file mode 100644
index 0000000..65ea498
--- /dev/null
+++ b/tests/test_client_compat_check.py
@@ -0,0 +1,135 @@
+from __future__ import annotations
+
+import json
+import os
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from scripts import client_compat_check
+
+
+class ClientCompatConfigTests(unittest.TestCase):
+    def _write_binary(self, path: Path, version: str) -> None:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(
+            "#!/bin/sh\n"
+            f"printf '%s\\n' 'reasonkit-think-mcp {version}'\n",
+            encoding="utf-8",
+        )
+        path.chmod(0o755)
+
+    def _write_cursor_config(self, home: Path, command: str) -> None:
+        path = home / ".cursor" / "mcp.json"
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(
+            json.dumps(
+                {
+                    "mcpServers": {
+                        "reasonkit-think": {
+                            "type": "stdio",
+                            "command": command,
+                            "args": [],
+                        }
+                    }
+                }
+            ),
+            encoding="utf-8",
+        )
+
+    def test_path_command_resolves_the_exact_candidate(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            home = root / "home"
+            binary = root / "bin" / "reasonkit-think-mcp"
+            self._write_binary(binary, "0.2.0")
+            self._write_cursor_config(home, "reasonkit-think-mcp")
+
+            with (
+                patch.object(client_compat_check.Path, "home", return_value=home),
+                patch.dict(os.environ, {"PATH": str(binary.parent)}),
+            ):
+                checks = client_compat_check.client_config_checks(root, binary)
+
+            cursor = next(check for check in checks if check["client"] == "cursor")
+            self.assertEqual(cursor["status"], "ok")
+            self.assertEqual(cursor["configured_command"], "reasonkit-think-mcp")
+            self.assertEqual(cursor["resolved_command"], str(binary.resolve()))
+            self.assertFalse(cursor["command_matches_release_binary"])
+            self.assertTrue(cursor["resolved_matches_release_binary"])
+            self.assertEqual(cursor["candidate_version"], "0.2.0")
+            self.assertEqual(cursor["resolved_version"], "0.2.0")
+            self.assertTrue(cursor["version_matches_release_binary"])
+
+    def test_path_install_with_candidate_version_is_accepted_but_not_exact(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            home = root / "home"
+            binary = root / "candidate" / "reasonkit-think-mcp"
+            installed = root / "path-bin" / "reasonkit-think-mcp"
+            self._write_binary(binary, "0.2.0")
+            self._write_binary(installed, "0.2.0")
+            self._write_cursor_config(home, "reasonkit-think-mcp")
+
+            with (
+                patch.object(client_compat_check.Path, "home", return_value=home),
+                patch.dict(os.environ, {"PATH": str(installed.parent)}),
+            ):
+                checks = client_compat_check.client_config_checks(root, binary)
+
+            cursor = next(check for check in checks if check["client"] == "cursor")
+            self.assertEqual(cursor["status"], "ok")
+            self.assertEqual(cursor["resolved_command"], str(installed.resolve()))
+            self.assertFalse(cursor["resolved_matches_release_binary"])
+            self.assertTrue(cursor["version_matches_release_binary"])
+
+    def test_path_install_with_an_old_version_reports_drift(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            home = root / "home"
+            binary = root / "candidate" / "reasonkit-think-mcp"
+            installed = root / "path-bin" / "reasonkit-think-mcp"
+            self._write_binary(binary, "0.2.0")
+            self._write_binary(installed, "0.1.1")
+            self._write_cursor_config(home, "reasonkit-think-mcp")
+
+            with (
+                patch.object(client_compat_check.Path, "home", return_value=home),
+                patch.dict(os.environ, {"PATH": str(installed.parent)}),
+            ):
+                checks = client_compat_check.client_config_checks(root, binary)
+
+            cursor = next(check for check in checks if check["client"] == "cursor")
+            self.assertEqual(cursor["status"], "drift")
+            self.assertEqual(cursor["resolved_command"], str(installed.resolve()))
+            self.assertEqual(cursor["candidate_version"], "0.2.0")
+            self.assertEqual(cursor["resolved_version"], "0.1.1")
+            self.assertFalse(cursor["version_matches_release_binary"])
+
+    def test_path_install_requires_reasonkit_version_identity(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            home = root / "home"
+            binary = root / "candidate" / "reasonkit-think-mcp"
+            installed = root / "path-bin" / "reasonkit-think-mcp"
+            self._write_binary(binary, "0.2.0")
+            installed.parent.mkdir(parents=True, exist_ok=True)
+            installed.write_text("#!/bin/sh\nprintf '%s\\n' '0.2.0'\n", encoding="utf-8")
+            installed.chmod(0o755)
+            self._write_cursor_config(home, "reasonkit-think-mcp")
+
+            with (
+                patch.object(client_compat_check.Path, "home", return_value=home),
+                patch.dict(os.environ, {"PATH": str(installed.parent)}),
+            ):
+                checks = client_compat_check.client_config_checks(root, binary)
+
+            cursor = next(check for check in checks if check["client"] == "cursor")
+            self.assertEqual(cursor["status"], "drift")
+            self.assertIsNone(cursor["resolved_version"])
+            self.assertFalse(cursor["version_matches_release_binary"])
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/test_community_contract.py b/tests/test_community_contract.py
new file mode 100644
index 0000000..511b518
--- /dev/null
+++ b/tests/test_community_contract.py
@@ -0,0 +1,78 @@
+from __future__ import annotations
+
+import json
+import re
+import unittest
+from collections import Counter
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parent.parent
+FEATURE_ID = re.compile(r"^(T\d+-[A-Z0-9]+)-")
+
+
+class CommunityContractTests(unittest.TestCase):
+    def test_public_contribution_and_security_paths_exist(self) -> None:
+        required = [
+            "CONTRIBUTING.md",
+            "SECURITY.md",
+            ".github/ISSUE_TEMPLATE/bug_report.yml",
+            ".github/ISSUE_TEMPLATE/feature_request.yml",
+            ".github/ISSUE_TEMPLATE/config.yml",
+            ".github/pull_request_template.md",
+        ]
+        for relative in required:
+            path = ROOT / relative
+            self.assertTrue(path.is_file(), relative)
+            self.assertTrue(path.read_text(encoding="utf-8").strip(), relative)
+
+        contributing = (ROOT / "CONTRIBUTING.md").read_text(encoding="utf-8")
+        self.assertIn("just check", contributing)
+        self.assertIn("evals/tool-packs.json", contributing)
+        self.assertIn("Conventional Commits", contributing)
+
+        security = (ROOT / "SECURITY.md").read_text(encoding="utf-8")
+        self.assertIn("privately", security)
+        self.assertNotIn("public issue", security.lower())
+
+    def test_feature_record_ids_are_unique_and_match_headings(self) -> None:
+        records = sorted((ROOT / "features").glob("*/*.md"))
+        ids: list[str] = []
+        for path in records:
+            match = FEATURE_ID.match(path.name)
+            if match is None:
+                continue
+            feature_id = match.group(1)
+            heading = path.read_text(encoding="utf-8").splitlines()[0]
+            self.assertTrue(heading.startswith(f"# {feature_id}:"), path.name)
+            ids.append(feature_id)
+
+        self.assertTrue(ids)
+        duplicates = [name for name, count in Counter(ids).items() if count > 1]
+        self.assertEqual(duplicates, [])
+
+    def test_feature_manifest_distinguishes_plans_from_runtime_truth(self) -> None:
+        source = (ROOT / "features" / "README.md").read_text(encoding="utf-8")
+        self.assertIn("planning records", source)
+        self.assertIn("evals/tool-packs.json", source)
+        self.assertIn("Partial overlap", source)
+        self.assertIn("T2-2F", source)
+
+        payload = json.loads(
+            (ROOT / "features" / "manifest.json").read_text(encoding="utf-8")
+        )
+        records = payload["records"]
+        manifest_paths = {record["path"] for record in records}
+        disk_paths = {
+            path.relative_to(ROOT).as_posix()
+            for path in (ROOT / "features").glob("*/*.md")
+            if FEATURE_ID.match(path.name)
+        }
+        self.assertEqual(manifest_paths, disk_paths)
+        self.assertEqual(len(records), len({record["id"] for record in records}))
+        for record in records:
+            self.assertEqual(record["record_state"], Path(record["path"]).parent.name)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/test_distribution_docs.py b/tests/test_distribution_docs.py
new file mode 100644
index 0000000..b564418
--- /dev/null
+++ b/tests/test_distribution_docs.py
@@ -0,0 +1,173 @@
+from __future__ import annotations
+
+import json
+import shutil
+import subprocess
+import tomllib
+import unittest
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parent.parent
+
+
+class DistributionDocsTests(unittest.TestCase):
+    def test_release_metadata_and_crate_boundary_are_synchronized(self) -> None:
+        cargo = tomllib.loads((ROOT / "Cargo.toml").read_text(encoding="utf-8"))
+        registry = json.loads((ROOT / "server.json").read_text(encoding="utf-8"))
+        package = cargo["package"]
+
+        self.assertEqual(package["version"], "0.2.0")
+        self.assertEqual(package["version"], registry["version"])
+        self.assertNotIn("exclude", package)
+        self.assertEqual(
+            package["include"],
+            [
+                "/src/**",
+                "/Cargo.toml",
+                "/Cargo.lock",
+                "/README.md",
+                "/CHANGELOG.md",
+                "/LICENSE",
+            ],
+        )
+
+    @unittest.skipUnless(shutil.which("cargo"), "cargo is not installed")
+    def test_packaged_file_set_excludes_internal_project_material(self) -> None:
+        result = subprocess.run(
+            ["cargo", "package", "--list", "--allow-dirty", "--locked"],
+            cwd=ROOT,
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(result.returncode, 0, result.stderr)
+
+        allowed_root_files = {
+            ".cargo_vcs_info.json",
+            "CHANGELOG.md",
+            "Cargo.lock",
+            "Cargo.toml",
+            "Cargo.toml.orig",
+            "LICENSE",
+            "README.md",
+        }
+        unexpected = [
+            path
+            for path in result.stdout.splitlines()
+            if path not in allowed_root_files and not path.startswith("src/")
+        ]
+        self.assertEqual(unexpected, [])
+
+    def test_readme_has_truthful_path_first_golden_path(self) -> None:
+        readme = (ROOT / "README.md").read_text(encoding="utf-8")
+        self.assertIn("## Quick start", readme)
+        self.assertNotIn("Five-minute path", readme)
+        self.assertNotIn("tracked warnings stay visible", readme)
+        self.assertIn("deny vulnerabilities and advisory warnings", readme)
+        self.assertIn(
+            "mcp-name: io.github.reasonkit/reasonkit-think",
+            readme,
+        )
+        self.assertIn("crates.io/crates/reasonkit-think-mcp", readme)
+        self.assertNotIn('crates.io/crates/reasonkit-think"', readme)
+        self.assertNotIn("ABSOLUTE PATH", readme)
+        self.assertIn('"command": "reasonkit-think-mcp"', readme)
+        self.assertIn("compatibility escape hatch", readme)
+        self.assertIn("not yet a live Registry record", readme)
+        self.assertIn("cargo install --locked --path .", readme)
+        self.assertIn("after v0.2.0 publication", readme)
+        self.assertIn(
+            "cargo install --locked --version 0.2.0 reasonkit-think-mcp",
+            readme,
+        )
+
+        sequence = [
+            "run_thinking_mode",
+            "add_thought_node",
+            "verify_thoughts",
+            "run_reasoning_checkpoint",
+            "consensus_answer",
+            "export_reasoning_audit",
+        ]
+        positions = [readme.index(name) for name in sequence]
+        self.assertEqual(positions, sorted(positions))
+
+    def test_client_docs_and_examples_use_path_commands(self) -> None:
+        client_docs = list((ROOT / "docs" / "clients").glob("*.md"))
+        self.assertTrue(client_docs)
+        for path in client_docs:
+            source = path.read_text(encoding="utf-8")
+            self.assertNotIn("/ABSOLUTE/PATH", source, path.name)
+            self.assertIn("reasonkit-think-mcp", source, path.name)
+
+        configs = list((ROOT / "examples" / "client-configs").glob("*.json"))
+        self.assertTrue(configs)
+        for path in configs:
+            payload = json.loads(path.read_text(encoding="utf-8"))
+            rendered = json.dumps(payload)
+            self.assertNotIn("/absolute/path", rendered, path.name)
+            self.assertIn("reasonkit-think-mcp", rendered, path.name)
+
+    def test_registry_manifest_targets_the_release_candidate(self) -> None:
+        path = ROOT / "server.json"
+        if not path.is_file():
+            self.fail("server.json must exist")
+        manifest = json.loads(path.read_text(encoding="utf-8"))
+        self.assertEqual(
+            manifest["$schema"],
+            "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
+        )
+        self.assertEqual(manifest["name"], "io.github.reasonkit/reasonkit-think")
+        self.assertEqual(manifest["version"], "0.2.0")
+        self.assertEqual(
+            manifest["packages"],
+            [
+                {
+                    "registryType": "cargo",
+                    "identifier": "reasonkit-think-mcp",
+                    "version": "0.2.0",
+                    "transport": {"type": "stdio"},
+                }
+            ],
+        )
+
+    def test_cargo_dist_bootstrap_is_config_only(self) -> None:
+        path = ROOT / "dist-workspace.toml"
+        if not path.is_file():
+            self.fail("dist-workspace.toml must exist")
+        config = tomllib.loads(path.read_text(encoding="utf-8"))
+        self.assertEqual(config["dist"]["cargo-dist-version"], "0.32.0")
+        self.assertEqual(config["dist"]["ci"], [])
+        self.assertEqual(config["dist"]["hosting"], ["github"])
+        self.assertFalse((ROOT / ".github" / "workflows" / "release.yml").exists())
+
+    @unittest.skipUnless(shutil.which("dist"), "cargo-dist is not installed")
+    def test_cargo_dist_manifest_needs_no_generated_workflow(self) -> None:
+        result = subprocess.run(
+            [
+                "dist",
+                "manifest",
+                "--artifacts=local",
+                "--output-format=json",
+                "--no-local-paths",
+            ],
+            cwd=ROOT,
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(result.returncode, 0, result.stderr)
+        self.assertFalse((ROOT / ".github" / "workflows" / "release.yml").exists())
+
+    def test_python_test_caches_are_ignored_repo_wide(self) -> None:
+        patterns = {
+            line.strip()
+            for line in (ROOT / ".gitignore").read_text(encoding="utf-8").splitlines()
+        }
+        self.assertIn("**/__pycache__/", patterns)
+        self.assertIn("*.py[cod]", patterns)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/test_mcp_contract.py b/tests/test_mcp_contract.py
new file mode 100644
index 0000000..16978c7
--- /dev/null
+++ b/tests/test_mcp_contract.py
@@ -0,0 +1,292 @@
+from __future__ import annotations
+
+import importlib
+import importlib.util
+import unittest
+from pathlib import Path
+
+
+MODULE_NAME = "evals.mcp_contract"
+try:
+    MODULE_SPEC = importlib.util.find_spec(MODULE_NAME)
+except ModuleNotFoundError:
+    MODULE_SPEC = None
+
+
+class ContractModulePresenceTests(unittest.TestCase):
+    def test_contract_module_exists(self) -> None:
+        self.assertIsNotNone(
+            MODULE_SPEC,
+            "evals.mcp_contract must provide the executable MCP contract",
+        )
+
+    def test_contract_eval_cli_exists(self) -> None:
+        self.assertTrue(
+            Path("evals/run_contract_evals.py").is_file(),
+            "evals/run_contract_evals.py must execute live contract evaluations",
+        )
+
+
+@unittest.skipUnless(MODULE_SPEC, "contract module is the TDD implementation target")
+class ContractTests(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        cls.contracts = importlib.import_module(MODULE_NAME)
+        cls.contract = cls.contracts.load_contract()
+
+    def test_tool_pack_contract_is_exact_nested_and_sorted(self) -> None:
+        self.assertEqual(self.contract["defaultToolPack"], "core")
+        self.assertEqual(
+            {name: len(tools) for name, tools in self.contract["toolPacks"].items()},
+            {"core": 13, "standard": 33, "full": 47},
+        )
+        self.assertEqual(self.contracts.validate_contract(self.contract), [])
+
+    def test_protocol_matrix_includes_current_and_legacy_versions(self) -> None:
+        self.assertEqual(
+            self.contract["protocolVersions"],
+            [
+                "2024-11-05",
+                "2025-03-26",
+                "2025-06-18",
+                "2025-11-25",
+                "2026-07-28",
+            ],
+        )
+
+    def test_tool_surface_reports_missing_unexpected_and_order_drift(self) -> None:
+        core = self.contract["toolPacks"]["core"]
+        actual = [{"name": name} for name in core]
+        self.assertEqual(
+            self.contracts.validate_tool_surface(self.contract, "core", actual),
+            [],
+        )
+
+        drifted = actual[1:] + [{"name": "unexpected_tool"}]
+        issues = self.contracts.validate_tool_surface(self.contract, "core", drifted)
+        self.assertTrue(any("missing" in issue for issue in issues))
+        self.assertTrue(any("unexpected" in issue for issue in issues))
+        self.assertTrue(any("lexical" in issue for issue in issues))
+
+    def test_schema_validator_rejects_untyped_composed_shapes(self) -> None:
+        invalid = {
+            "properties": {
+                "mode": {"enum": ["quick", "test"]},
+                "limit": {"type": "integer"},
+            }
+        }
+        issues = self.contracts.schema_type_issues(invalid)
+        self.assertIn("$: missing type", issues)
+        self.assertIn("$.properties.mode: missing type", issues)
+        self.assertNotIn("$.properties.limit: missing type", issues)
+
+    def test_fail_closed_evaluation_accepts_only_blocked_outcomes(self) -> None:
+        evaluation = self.contracts.evaluate_fail_closed(
+            evidence_free_update={
+                "result": {
+                    "isError": False,
+                    "structuredContent": {
+                        "code": "invalid_input",
+                        "assumption": {
+                            "status": "unresolved",
+                            "criticality": "critical",
+                        },
+                        "blocking_assumption_ids": ["assumption-1"],
+                    },
+                    "content": [{"type": "text", "text": "evidence required"}],
+                }
+            },
+            criticality_downgrade={
+                "result": {
+                    "structuredContent": {
+                        "verification_matrix": [
+                            {
+                                "claim": "Release requires approval",
+                                "critical": True,
+                                "status": "data_deficit",
+                            }
+                        ]
+                    }
+                }
+            },
+            checkpoint={
+                "result": {
+                    "content": [
+                        {
+                            "type": "text",
+                            "text": '{"checkpoint":{"route_decision":"GATHER_MORE_EVIDENCE","blocking_gaps":["critical"]}}',
+                        }
+                    ]
+                }
+            },
+            consensus={
+                "result": {
+                    "isError": True,
+                    "structuredContent": {"code": "consensus_blocked"},
+                    "content": [{"type": "text", "text": "blocked"}],
+                }
+            },
+            unknown_stage={
+                "result": {
+                    "isError": True,
+                    "structuredContent": {"code": "invalid_input"},
+                    "content": [{"type": "text", "text": "unknown stage"}],
+                }
+            },
+            pipeline_override={"result": {"structuredContent": {"ok": True}}},
+            post_pipeline_audit={
+                "result": {
+                    "structuredContent": {
+                        "payload": {
+                            "verification_policy": {
+                                "min_independent_groups": 3,
+                                "allow_tier3_for_independence": False,
+                                "require_tier1_unambiguous_for_critical": True,
+                                "fail_closed_on_critical_unresolved": True,
+                            }
+                        }
+                    }
+                }
+            },
+            scaffold_run={
+                "result": {
+                    "structuredContent": {
+                        "policy_blocked": True,
+                        "route_decision": "GATHER_MORE_EVIDENCE",
+                        "pipeline_route": "GATHER_MORE_EVIDENCE",
+                        "graph_authorship": {
+                            "agent_authored": 0,
+                            "scaffold_nodes": 3,
+                            "frontier_scaffold_only": True,
+                        },
+                    }
+                }
+            },
+        )
+        self.assertTrue(evaluation["passed"])
+        self.assertTrue(all(evaluation["checks"].values()))
+
+    def test_fail_closed_evaluation_catches_a_bypass(self) -> None:
+        bypass = {
+            "result": {
+                "isError": False,
+                "content": [{"type": "text", "text": '{"status":"verified"}'}],
+            }
+        }
+        evaluation = self.contracts.evaluate_fail_closed(
+            evidence_free_update=bypass,
+            criticality_downgrade={
+                "result": {
+                    "structuredContent": {
+                        "verification_matrix": [
+                            {
+                                "claim": "Release requires approval",
+                                "critical": False,
+                                "status": "VERIFIED",
+                            }
+                        ]
+                    }
+                }
+            },
+            checkpoint={
+                "result": {
+                    "content": [
+                        {
+                            "type": "text",
+                            "text": '{"checkpoint":{"route_decision":"PROCEED_WITH_CAVEATS","blocking_gaps":[]}}',
+                        }
+                    ]
+                }
+            },
+            consensus=bypass,
+            unknown_stage=bypass,
+            pipeline_override=bypass,
+            post_pipeline_audit={
+                "result": {
+                    "structuredContent": {
+                        "payload": {
+                            "verification_policy": {
+                                "min_independent_groups": 1,
+                                "allow_tier3_for_independence": True,
+                                "require_tier1_unambiguous_for_critical": False,
+                                "fail_closed_on_critical_unresolved": False,
+                            }
+                        }
+                    }
+                }
+            },
+            scaffold_run={
+                "result": {
+                    "structuredContent": {
+                        "policy_blocked": False,
+                        "route_decision": "PROCEED_WITH_CAVEATS",
+                        "pipeline_route": "GATHER_MORE_EVIDENCE",
+                        "graph_authorship": {
+                            "agent_authored": 0,
+                            "scaffold_nodes": 3,
+                            "frontier_scaffold_only": True,
+                        },
+                    }
+                }
+            },
+        )
+        self.assertFalse(evaluation["passed"])
+        self.assertFalse(any(evaluation["checks"].values()))
+
+    def test_runtime_probe_and_matrix_api_exists(self) -> None:
+        self.assertTrue(
+            hasattr(self.contracts, "validate_runtime_probe"),
+            "contract module must validate a live tools/list response",
+        )
+        self.assertTrue(
+            hasattr(self.contracts, "matrix_jobs"),
+            "contract module must generate every pack/protocol pairing",
+        )
+
+    def test_matrix_jobs_cover_all_fifteen_pack_protocol_pairs(self) -> None:
+        jobs = self.contracts.matrix_jobs(self.contract)
+        self.assertEqual(len(jobs), 15)
+        self.assertEqual(jobs[0], ("core", "2024-11-05"))
+        self.assertEqual(jobs[-1], ("full", "2026-07-28"))
+
+    def test_runtime_probe_requires_negotiation_exact_surface_and_typed_schemas(self) -> None:
+        protocol = "2026-07-28"
+        tools = [
+            {
+                "name": name,
+                "inputSchema": {"type": "object", "properties": {}},
+            }
+            for name in self.contract["toolPacks"]["core"]
+        ]
+        initialize = {
+            "result": {
+                "protocolVersion": protocol,
+                "serverInfo": {"name": "reasonkit-think-mcp"},
+            }
+        }
+        self.assertEqual(
+            self.contracts.validate_runtime_probe(
+                self.contract,
+                pack="core",
+                protocol=protocol,
+                initialize=initialize,
+                tools=tools,
+            ),
+            [],
+        )
+
+        initialize["result"]["serverInfo"]["name"] = "wrong-server"
+        tools[0]["inputSchema"] = {"properties": {"value": {"enum": ["x"]}}}
+        issues = self.contracts.validate_runtime_probe(
+            self.contract,
+            pack="core",
+            protocol=protocol,
+            initialize=initialize,
+            tools=tools,
+        )
+        self.assertTrue(any("server name" in issue for issue in issues))
+        self.assertTrue(any("missing type" in issue for issue in issues))
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/test_smoke_contract.py b/tests/test_smoke_contract.py
new file mode 100644
index 0000000..af5f416
--- /dev/null
+++ b/tests/test_smoke_contract.py
@@ -0,0 +1,14 @@
+from pathlib import Path
+import unittest
+
+
+class SmokeContractTests(unittest.TestCase):
+    def test_smoke_runs_full_pack_and_checks_exact_inventory(self) -> None:
+        source = Path("scripts/smoke_test.py").read_text(encoding="utf-8")
+        self.assertIn('env["REASONKIT_TOOL_PACK"] = "full"', source)
+        self.assertIn('root / "evals" / "tool-packs.json"', source)
+        self.assertIn("tools == expected_tools", source)
+
+
+if __name__ == "__main__":
+    unittest.main()