A governance and multi-tenancy gateway that fronts backends (agentmemory, issue trackers, a docs corpus, CI) behind the Model Context Protocol, with per-tenant isolation, confirmation-gated writes, and an audit trail.
Server has 17 tools with basic schemas and descriptions, but significant gaps limit production readiness. Tool naming follows verb_noun convention (ci.status, docs.search, issue.create), which is good. However, descriptions are minimal (10-50 chars typically), parameter descriptions lack detail on constraints/formats, and output schemas are not documented. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Error handling and recovery guidance are absent from descriptions. The memory.* tools and ci.* tools have the clearest definitions, but docs.* and issue.* lack specificity on return structures.
Get CI job build artifact
List CI job builds
Get CI job build log
Rerun a CI job build
Get CI job status
Stop a running CI job build
Get documentation content
List documentation repositories
Output schemas not documented. Tools return data but LLMs cannot plan downstream calls without knowing response structure (e.g., does docs.search return [title, url, snippet] or [id, content, metadata]?). This forces agents to guess or make discovery calls.
Parameter descriptions lack constraint details. 'job' param in ci.* tools has no guidance on format (alphanumeric? with slashes?). 'query' in docs.search and issue.search lacks examples of valid syntax. 'updates' in issue.update is vague ('Fields to update', which fields are allowed?). LLMs cannot validate inputs without explicit constraints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Search documentation corpus
Create a new issue
Get issue details
Search issues in issue tracker
Update an issue
Delete memory content
Get memory content
Set memory content
List available tools
No tool annotations despite clear risk levels. ci.rerun and ci.stop are WRITE operations; issue.create and issue.update are WRITE; memory.delete is DESTRUCTIVE. Annotations (readOnlyHint, destructiveHint, idempotentHint) are missing, preventing agents from understanding safety implications and planning confirmation gates.
Descriptions are too brief (10-50 chars). 'Get CI job status', 'List CI job builds', 'Search documentation corpus' lack context on WHEN to use each tool vs alternatives, WHAT the return value contains, and any prerequisites. Baseline for A+ tools is 50-200 chars with actionable guidance.
No error handling guidance in descriptions. Tools like ci.rerun, issue.create, and memory.delete can fail (job not found, invalid input, permission denied), but descriptions do not explain recovery steps. Agents need 'If X fails, try Y' guidance to self-correct.