MCP server for the m-dev-tools org catalog. Exposes three first-class agent tools:

• route_intent(query) — plain-English intent → typed IDs (e.g. "parse JSON in M" → module:m-stdlib#STDJSON)
• describe(typed_id) — typed ID → pointer-blob (manifest URL, AGENTS.md URL, verification commands, …)
• verify(repo) — list a repo's declared verification commands (does not execute them)

The server reads the catalog at https://github.com/m-dev-tools/.github over the network at call time; it is a thin wrapper, not a cache. See AGENTS.md for the contract and the AI users guide for the full walk-through.

mcp-name: io.github.m-dev-tools/m-dev-tools-mcp

pip install m-dev-tools-mcp
# or:
uvx m-dev-tools-mcp
# or from a GitHub Release wheel:
pip install https://github.com/m-dev-tools/m-dev-tools-mcp/releases/download/v0.2.4/m_dev_tools_mcp-0.2.4-py3-none-any.whl

Point any MCP client at the m-dev-tools-mcp binary the install provides:

{
  "mcpServers": {
    "m-dev-tools": { "command": "m-dev-tools-mcp" }
  }
}

Or for clients that consult the public MCP registry:

io.github.m-dev-tools/m-dev-tools-mcp

make install        # creates .venv and installs editable + dev deps
make test           # pytest
make check          # lint + mypy + test + check-manifest + check-agents
make build          # → wheel-out/m_dev_tools_mcp-<ver>-py3-none-any.whl

• Architecture: m-dev-tools/.github's docs/ai-discoverability/AI-discoverability-architecture.md
• Plan + phases: docs/ai-discoverability/phases/
• Release process: tag vX.Y.Z on main → .github/workflows/release.yml builds the wheel, attaches it to a GitHub Release, publishes to PyPI via Trusted Publisher OIDC, and updates the MCP registry record via GitHub OIDC.

AGPL-3.0. Same license as every other m-dev-tools repo.