Analyze Chess Games and Puzzles with Stockfish

ChessAgine MCP is a Model Context Protocol server that gives LLMs deep chess awareness by exposing real-time board state, Stockfish analysis, opening databases, Lichess games, and neural chess engines including Maia2, Leela, and Elite Leela. It renders individual positions and full PGN games for visual analysis, letting an AI agent reason about positions, evaluate variations, detect tactical or strategic themes, explore game databases, and interact directly with chess engines rather than reasoning about chess text alone.

Documented usage patterns include pulling a user's last Lichess game by username and analyzing it with Stockfish, comparing and contrasting what Stockfish thinks of a given FEN position against Leela and Maia's neural evaluations, and analyzing an opening repertoire from a chessboard. bash git clone https://github.com/jalpp/chessagine-mcp.git cd chessagine-mcp npm install npm run build

Use this connector when you want an AI assistant to analyze chess positions or games with real engine backing - comparing classical search-based evaluation (Stockfish) against neural, human-style engines (Maia2, Leela, Elite Leela) - or to pull and review Lichess game history and render boards visually rather than describing positions in text. It's suited to chess study, opening-repertoire review, and engine-comparison workflows. It is less useful outside of chess-specific analysis, and self-hosted deployment requires either Node.js 22+ locally or a Vercel deployment, so environments without either path cannot run it.

Accepts a FEN position, a Lichess username and game, or an opening line as a chess query, and returns engine evaluations from Stockfish and the neural engines (Maia2, Leela, Elite Leela), rendered board or PGN visualizations, and opening-database or game-history data for the AI agent to reason over.

Installs into Claude Desktop either as a prebuilt .mcpb extension file or via a local Node.js build (Node.js 22+) registered in claude_desktop_config.json, or can be self-hosted by forking the repository and deploying to Vercel with no environment variables required, exposing the server at https://your-project.vercel.app/mcp. For local development: npm run build:mcp builds the MCP server layer and generates the .mcpb file, npm run build:ui builds the ChessAgine UI HTML, npm run start starts the server, and npm run debug opens the MCP Inspector to verify changes. The project is licensed under the MIT/GPL License (the /remote component specifically under GPL).

Chess players and analysts who want an AI assistant to evaluate positions and games using real chess engines, compare classical versus neural engine assessments, review Lichess game history, or study opening repertoires with visual board rendering instead of text-only chess discussion.