feat(mcp): scope completion to authorized Assets and sanitize tool errors by kl3inIT · Pull Request #121 · kl3inIT/OrgMemory

kl3inIT · 2026-07-29T07:00:20Z

What

Closes the two MCP gaps that were reachable under the current stateless transport.

1. Permission-scoped completion (completion capability was false)

released_prompt and both Asset resource templates now answer completion/complete:

every suggestion comes from one authorized Asset delivery call for the current identity, so completion never reveals an Asset the caller could not already read;
an already-resolved argument narrows the next one, so release_id only offers releases of the chosen asset_id (CompleteRequest.context());
a prefix matches an identifier prefix or the Asset title/namespace/slug/version, so typing onboarding still finds the right release;
results cap at 20 with total/hasMore;
a delivery failure returns no suggestion instead of an error, so completion cannot become a second existence channel.

2. Sanitized tool errors

AbstractSyncMcpToolMethodCallback builds the returned error text as e.getMessage() + "\n" + rootCause.getMessage(). Our gateway messages were already sanitized, but the root cause was not: an UnknownHostException or ResourceAccessException published the internal API host and request URL to the MCP caller.

McpFailureBoundary now wraps every tool, prompt, and resource call: it logs the cause server-side and rethrows a cause-free failure carrying only the sanitized gateway message. The gateway clients keep their cause for diagnosis.

Deliberately not done

Blocked by protocol: STATELESS, each needing a separate transport decision rather than a quiet change here:

per-actor filtering of tools/list / prompts/list / resources/list — the stateless annotation factory registers specifications once at startup;
progress notifications, logging notifications, elicitation, sampling — the stateless server opens no server-to-client channel.

Docs

docs/specs/domains/assistant-and-mcp.md was stale on two counts and is corrected against the code: it claimed the surface was one search_knowledge tool (it is nine tools, two resource templates, one prompt) and that the inbound bearer was forwarded downstream (McpApiAuthorization exchanges it and never forwards).

Verification

gradlew :apps:mcp:test — 14 classes / 46 tests, 0 failures. New: 7 completion tests, 1 error-surface regression test that fails without the boundary fix; OrgMemoryMcpContextTests (a @SpringBootTest) proves the enabled capability wires three completion specifications.
gradlew compileJava (whole repo) — pass.
Static-analysis mechanical floor — pass. JetBrains inspection unavailable in this session; used the documented fallback.
Full clean test not run: the change is confined to apps/mcp, a leaf app no module depends on.

Known limits

Completion values are UUIDs, because MCP inserts the literal value into the argument. A human-readable coordinate would require changing the public prompt contract.
release_id suggests only each Asset current release; /api/asset-delivery exposes no release history.

🤖 Generated with Claude Code

Summary by CodeRabbit

New Features
- Enabled MCP completion suggestions for authorized prompt arguments and asset resource templates.
- Added filtering, prefix matching, asset scoping, pagination, and support for asset and release URI completions.
Bug Fixes
- Sanitized downstream tool, resource, prompt, and search errors to prevent internal transport details from appearing in responses.
Documentation
- Clarified completion behavior and failure handling in the architecture documentation.

…rors Publish MCP completion for the released Prompt argument and both Asset resource templates. Every suggestion comes from one authorized Asset delivery call for the current identity, an already-resolved argument narrows the next one, results are capped with total/hasMore, and a delivery failure yields no suggestion instead of an error so completion cannot become a second existence channel. Keep the downstream cause off the MCP boundary. The annotation runtime appends the root cause message to the tool error it returns, so an unresolved host or request URL reached the caller. The boundary now logs the cause and rethrows a cause-free failure carrying only the already sanitized gateway message. Reconcile the assistant-and-mcp spec and test matrix, which still claimed the surface was one search tool and that the inbound bearer was forwarded downstream. Verified with gradlew :apps:mcp:test (46 tests) and gradlew compileJava. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

…mmit The spec and test matrix now describe the completion surface and the sanitized error boundary introduced in 2f6b0f8, so the reconciliation marker moves to that commit. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

coderabbitai · 2026-07-29T07:00:38Z

Warning

Review limit reached

@kl3inIT, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 32 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 3f878027-7575-42e7-9b07-5fd640e83feb

📥 Commits

Reviewing files that changed from the base of the PR and between 7c809f5 and ab028aa.

⛔ Files ignored due to path filters (1)

docs/specs/domains/assistant-and-mcp.md is excluded by !docs/**

📒 Files selected for processing (3)

ARCHITECTURE.md
apps/mcp/src/main/java/com/orgmemory/mcp/McpFailureBoundary.java
apps/mcp/src/test/java/com/orgmemory/mcp/McpFailureBoundaryTests.java

📝 Walkthrough

Walkthrough

Adds MCP completion support for authorized asset and release values, enables completion capabilities, and introduces sanitized cause-free failure handling across asset, knowledge, resource, and prompt operations with corresponding tests and architecture documentation.

Changes

MCP completion and failure handling

Layer / File(s)	Summary
Authorized completion flow `apps/mcp/src/main/java/com/orgmemory/mcp/AssetCompletionAdapter.java`, `apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryResources.java`, `apps/mcp/src/main/resources/application.yml`	Adds authorized prompt and resource completions with filtering, scoping, paging, and completion capability enablement.
Sanitized downstream failures `apps/mcp/src/main/java/com/orgmemory/mcp/McpFailureBoundary.java`, `apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryTools.java`, `apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryResources.java`, `apps/mcp/src/main/java/com/orgmemory/mcp/KnowledgeSearchTool.java`, `apps/mcp/src/main/java/com/orgmemory/mcp/ReleasedPromptAdapter.java`	Routes downstream gateway calls through a failure boundary that logs sanitized messages and throws cause-free MCP failures.
Completion and error-surface validation `apps/mcp/src/test/java/com/orgmemory/mcp/AssetCompletionAdapterTests.java`, `apps/mcp/src/test/java/com/orgmemory/mcp/McpToolErrorSurfaceTests.java`, `apps/mcp/src/test/java/com/orgmemory/mcp/OrgMemoryMcpContextTests.java`, `ARCHITECTURE.md`	Tests completion filtering, scoping, paging, published references, and sanitized tool errors, and documents the behavior.

Estimated code review effort: 4 (Complex) | ~45 minutes

Sequence Diagram(s)

sequenceDiagram
  participant MCPClient
  participant AssetCompletionAdapter
  participant McpApiAuthorization
  participant AssetDeliveryApiClient
  MCPClient->>AssetCompletionAdapter: Request completion for an argument
  AssetCompletionAdapter->>McpApiAuthorization: Resolve authorized identity
  McpApiAuthorization-->>AssetCompletionAdapter: Authorized context
  AssetCompletionAdapter->>AssetDeliveryApiClient: Search readable assets
  AssetDeliveryApiClient-->>AssetCompletionAdapter: Asset and release summaries
  AssetCompletionAdapter-->>MCPClient: Filtered paged completions

Possibly related PRs

kl3inIT/OrgMemory#60: Directly overlaps the MCP asset delivery resources and tools updated by this PR.
kl3inIT/OrgMemory#61: Introduces the authorized asset-delivery surface used by these resource and tool changes.
kl3inIT/OrgMemory#88: Introduces the skill distribution tools now covered by sanitized failure handling.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main change: scoped MCP completion and sanitized tool failures.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch feat/mcp-scoped-completion

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

kl3inIT · 2026-07-29T10:22:13Z

@CodeRabbit review

coderabbitai · 2026-07-29T10:22:17Z

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

coderabbitai

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@apps/mcp/src/main/java/com/orgmemory/mcp/AssetCompletionAdapter.java`:
- Around line 138-150: Update the completion flow so authorized() receives the
typed prefix and passes it as the query argument to assets.search when
non-empty, while retaining null for an empty prefix. Verify
AssetDeliveryApiClient.search supports prefix narrowing without removing the
existing title/namespace/slug/versionLabel filtering performed by matches() or
breaking AssetCompletionAdapterTests.

In `@apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryResources.java`:
- Around line 37-58: Update getAsset and getAssetRelease so
JsonProcessingException is caught inside each lambda before invoking
McpFailureBoundary.sanitized. Convert the checked serialization failure into
AssetDeliveryGatewayException, while preserving the existing authorization,
asset lookup, release lookup, and sanitized boundary behavior.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: c0aa0a7a-064d-4787-bae6-c3ca8b21c680

📥 Commits

Reviewing files that changed from the base of the PR and between d4d6dc3 and 7c809f5.

⛔ Files ignored due to path filters (2)

docs/specs/domains/assistant-and-mcp.md is excluded by !docs/**
docs/tests/domains/assistant-and-mcp.md is excluded by !docs/**

📒 Files selected for processing (11)

ARCHITECTURE.md
apps/mcp/src/main/java/com/orgmemory/mcp/AssetCompletionAdapter.java
apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryResources.java
apps/mcp/src/main/java/com/orgmemory/mcp/AssetDeliveryTools.java
apps/mcp/src/main/java/com/orgmemory/mcp/KnowledgeSearchTool.java
apps/mcp/src/main/java/com/orgmemory/mcp/McpFailureBoundary.java
apps/mcp/src/main/java/com/orgmemory/mcp/ReleasedPromptAdapter.java
apps/mcp/src/main/resources/application.yml
apps/mcp/src/test/java/com/orgmemory/mcp/AssetCompletionAdapterTests.java
apps/mcp/src/test/java/com/orgmemory/mcp/McpToolErrorSurfaceTests.java
apps/mcp/src/test/java/com/orgmemory/mcp/OrgMemoryMcpContextTests.java

📜 Review details

🧰 Additional context used

📓 Path-based instructions (5)

**/*