Chess MCP Server

A Model Context Protocol (MCP) server that provides a chess-playing environment backed by the Stockfish engine. Designed for use with LLM clients (like Claude Desktop) to ensure all chess moves are legal and engine-calculated.

Features

• Legal Move Validation: All moves are validated against chess rules
• Stockfish Integration: Uses the Stockfish chess engine for AI moves
• MCP Protocol: Communicates via JSON-RPC over stdio
• In-Memory Game State: Maintains multiple concurrent games

Installation

pip install chess-mcp

You'll also need to install Stockfish:

• Windows: Download from Stockfish website
• macOS: brew install stockfish
• Linux: sudo apt install stockfish (or equivalent)

Usage

1. Set the STOCKFISH_PATH environment variable to the path of your Stockfish executable:

   export STOCKFISH_PATH=/path/to/stockfish
   

2. Run the server:

   python -m chess_mcp
   

MCP Tools

new_game()

Starts a new chess game from the standard initial position.

Returns:

• game_id: Unique identifier for the game
• fen: Current position in FEN notation
• legal_moves: List of all legal moves in UCI format

make_move(game_id, move_uci)

Applies a human move to the game. The move must be in UCI format (e.g., "e2e4").

Parameters:

• game_id: Identifier of the active game
• move_uci: Move in UCI format

Returns:

• ok: Whether the move was legal and applied
• fen: Updated board position (if move was legal)
• legal_moves: Updated list of legal moves

engine_move(game_id, depth=14)

Asks Stockfish to calculate and play one move for the side to move.

Parameters:

• game_id: Identifier of the active game
• depth: Search depth for Stockfish (higher = stronger but slower)

Returns:

• engine_move: The move played by Stockfish in UCI format
• fen: Updated board position after the move
• legal_moves: Updated list of legal moves

Development

To set up for development:

git clone https://github.com/yourusername/chess-mcp.git
cd chess-mcp
pip install -e .

License

MIT License