apm-mcp-server: アジャイルプラクティスマップ MCPサーバー

What is apm-mcp-server?

apm-mcp-server is an MCP server that lets large language models query the Agile Practice Map knowledge base. It serves users who need programmatic access to agile practice information through natural language questions or list retrieval.

How to use apm-mcp-server?

Install Python and the uv package manager, then clone the repository and set an OpenAI API key in a .env file. Configure your MCP client (e.g., Claude Desktop) to run the server with uv run -m apm_mcp_server or via Docker. Use the query_apm tool to ask questions and list_apm_practices to get all practice names. A pre-built Chroma DB vector database is included for immediate use.

Key features of apm-mcp-server

• Two MCP tools: query_apm and list_apm_practices
• Resource to retrieve a practice’s URL by name
• Built‑in Chroma DB vector database with Agile Practice Map data
• Ready‑to‑run from cloned repository with uv
• Docker deployment option for containerized environments
• Included populate_db script to build custom knowledge bases
• MIT licensed – free to use, modify, and distribute

Use cases of apm-mcp-server

• Ask natural‑language questions about specific agile practices and get answers from the map.
• Retrieve the full list of agile practice names for documentation or planning.
• Look up the web URL of a given practice directly via an MCP resource.
• Extend the knowledge base by adding new web pages or text files with the populate_db tool.

FAQ from apm-mcp-server

What data does apm-mcp-server use?

It uses the Agile Practice Map knowledge base (https://www.agile-studio.jp/agile-practice-map) stored as a local Chroma DB vector database in the data/ directory.

How do I install and run apm-mcp-server?

Install uv, clone the repository, set OPENAI_API_KEY in a .env file, then configure your MCP client to run uv run -m apm_mcp_server from the project directory. Docker build instructions are also provided.

Does apm-mcp-server require external APIs?

Yes, it requires an OpenAI API key for generating answers via the large language model. The key must be set in the .env file or passed as an environment variable in the Docker configuration.

Can I use apm-mcp-server with my own knowledge content?

Yes. The included populate_db.py script can build a vector database from a web page URL or a text file, allowing you to add custom knowledge to the server.

What transport protocol does apm-mcp-server use?

The server uses stdio transport, as configured in MCP client json files (e.g., claude_desktop_config.json). Docker mode also relies on stdio communication.