mcp-project-context
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
mcp-project-context
Status: work in progress. Built ticket by ticket with an agentic workflow, see How this was built.
An MCP server that gives coding agents such as Claude Code safe, read-only access to a repository's project knowledge: architecture decision records (ADRs), tickets and team conventions.
Why
Coding agents guess when they lack context. They re-open settled decisions, ignore conventions and miss acceptance criteria. Pasting docs into every prompt is tedious and leaks whatever happens to be in them.
This server lets the agent look things up itself, with guardrails:
Read-only by default. Write tools exist only when explicitly enabled.
Stays inside the project root. Path traversal and symlink escapes are blocked.
Redacts before it returns. Secrets and personal data are masked at the output boundary.
Logs to stderr only, with an audit line per call and no file contents.
Related MCP server: Gumi-MCP
Tools
Tool | What it does |
| List ADRs, optionally filtered by status |
| Read one ADR as structured sections |
| List local tickets, filter by status or label |
| Read one ticket with its acceptance criteria |
| Return CONVENTIONS.md, CLAUDE.md, CONTRIBUTING.md |
| Keyword search across all of the above, with snippets |
| Write a new ADR draft (only with |
ADRs and tickets are also exposed as resources: adr://{id}, ticket://{id}.
Quickstart
pnpm install
pnpm dev # serves examples/demo-project over stdioUse it in Claude Code (from your project directory):
claude mcp add project-context -- npx mcp-project-context --root .Configuration
Flag | Default | Meaning |
| required | Project root. Nothing outside it is readable. |
|
| Where ADRs live, relative to root |
|
| Where ticket markdown files live |
|
| Larger files are refused |
| off | Registers |
Demo
TODO (ticket 10): GIF of Claude Code using the server on the demo project.
How this was built
TODO (ticket 10): Short summary of the workflow and numbers from docs/WORKLOG.md:
total time, number of tickets, questions the agent asked, mistakes caught in review.
Requirements: docs/requirements.md ·
Architecture: docs/architecture.md ·
Decisions: docs/adr/ · Log: docs/WORKLOG.md
Limitations
TODO: be honest here. For example: redaction is pattern-based and will miss some secrets; search is keyword-based, not semantic.
License
MIT
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Run verified read-only code tools: quant diagnostics + agent-ops preflight, no source exposure.
Read-only AI coding tools for change verification, release readiness, capacity, and guidance.
Read-only tools over the Safer Agentic AI framework: 239 patterns + 15 heuristics.
Context and control for any AI. Agent identity, scoped access, and proof across tools.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceProvides AI agents with secure, read-only file system access to analyze and understand project codebases, enabling multi-repository context aggregation and cross-project code tracing.5MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to safely inspect a local repository's code and metadata while blocking private data from leaving the machine, providing read-only tools for search, change tracking, and integrity verification.MIT
- FlicenseNot gradedqualityBmaintenanceEnables AI coding agents to retrieve focused, bounded repository context locally and read-only, reducing token usage while never modifying the repository or executing its code.-
- AlicenseAqualityAmaintenanceEnables a coding agent to get short, token-efficient answers about a project through grouped searches, focused symbol or section reads, and bounded read-only script output, with safety limits on paths, secrets, commands, and response size.3MIT