Install from Source
Install the open-source Clean MCP server on your own machine.
The open-source edition (github.com/cleanmcp/clean-mcp, MIT) runs entirely on your machine. Embeddings, metadata, and source files live in ~/.clean/ — nothing leaves your computer.
Requirements
- Python 3.10–3.13
- git (only needed to clone GitHub repos for indexing)
Install
This project uses a src/ layout, so the module command only works after the package is installed into the Python environment you're using. If you create or activate a virtualenv, run python -m pip install -e ".[dev]" again inside it before starting the server. The dev extra installs the project's local tooling, including pytest and ruff.
The first search or index downloads the embedding model all-MiniLM-L6-v2 (~90 MB) into ~/.cache/huggingface/ if it isn't already cached. After that, the model is reused from disk and doesn't download again.
Heavy dependencies load lazily
The server starts with heavy dependencies lazy-loaded. Launching clean reaches Starting Clean MCP server quickly; the embedding model and LanceDB stack load only when you first call search_code or index_repo in that process.
Next steps
- Run the server and wire it into your agent
- Configure data locations and the embedding model
- Connect your agents