ChessAgine (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 engines including Maia2, Leela, and Elite Leela.

It also renders individual positions and full PGN games for in-depth visual analysis—enabling AI agents to reason about positions, evaluate variations, detect themes, explore game databases, and interact directly with chess engines.

Preview

Installation

ChessAgine MCP is aimed at chess players as much as developers — if you're not sure whether you have Node.js/npm installed, see the prerequisites note in install.md before picking an option below.

Option 1: Config File (Recommended, most reliable)

This is currently the most stable way to connect ChessAgine MCP to Claude Desktop.

Prerequisites

• Node.js 22+
• npm or yarn package manager
• Anthropic MCP Bundle CLI (npm install -g @anthropic-ai/mcpb)

Clone and Setup

git clone https://github.com/jalpp/chessagine-mcp.git
cd chessagine-mcp
npm install
npm run build

Configure Claude Desktop

MacOs

Add to ~/Library/Application Support/Claude/claude_desktop_config.json

Windows

Run (Get-AppxPackage *Claude*).PackageFamilyName to find the virtualized Claude Desktop folder Add to %LOCALAPPDATA%\Packages\Claude_virtualized_folder\LocalCache\Roaming\Claude\claude_desktop_config.json

Linux

Add to ~/.config/Claude/claude_desktop_config.json

macOS/Linux:

{
  "mcpServers": {
    "chessagine-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/chessagine-mcp/build/runner/stdio.js"],
      "env": {
        "LICHESS_API_TOKEN": "optional-your-lichess-api-token",
        "CHESSBOARD_MAGIC_PAT": "optional-your-chessboardmagic-pat",
        "POSIRA_API_KEY": "optional-your-posira-api-key"
      }
    }
  }
}

Windows:

{
  "mcpServers": {
    "chessagine-mcp": {
      "command": "node", 
      "args": ["C:\\absolute\\path\\to\\chessagine-mcp\\build\\runner\\stdio.js"],
      "env": {
        "LICHESS_API_TOKEN": "optional-your-lichess-api-token",
        "CHESSBOARD_MAGIC_PAT": "optional-your-chessboardmagic-pat",
        "POSIRA_API_KEY": "optional-your-posira-api-key"
      }
    }
  }
}

Option 2: Using MCPB File (experimental — currently less stable)

[!WARNING]
The .mcpb install path has known rough edges in some Claude Desktop versions. If it fails to load, use Option 1 above instead. Its one advantage: Claude Desktop runs it with its own bundled Node runtime, so you don't need Node.js/npm installed yourself for this option.

Download the chessagine-mcp.mcpb file and install it directly in Claude Desktop:

1. Download the latest release from GitHub releases
2. Open Claude Desktop
3. Go to Settings → Extensions → Install from file
4. Select the chessagine-mcp.mcpb file
5. Restart Claude Desktop

[!NOTE]
To make sure its working correctly ask it to render the chessboard or a specific chess query

Option 3: Connect to the Hosted Remote Server (No Install)

Skip the build entirely and point any MCP-compatible client at the hosted deployment:

https://chessagine-mcp.vercel.app/mcp

This is a Streamable HTTP endpoint — no API key or local setup required. Exact steps differ by client (config file vs. UI), so see install.md for copy-pasteable setup steps covering Claude Desktop, Claude Code, LibreChat, Open WebUI, AnythingLLM, Jan, Goose, Cursor, Windsurf, Cline, VS Code, Continue.dev, Zed, and other MCP-compatible GUIs — both open source and proprietary — for both the remote URL and the local stdio server.

Usage:

• show me my last Lichess game I played, I'm insert_your_username there, also analyze the game using Stockfish
• given fen compare and constrast what stockfish thinks vs Leela and Maia
• analyze my opening rep from Chessboard magic.

ChessAgine.Skill

to properly use ChessAgine MCP, give LLM access to how to properly use the it via .skill file in ./chessagine-skill/ folder

Deploy your own instance

You can deploy your own copy to Vercel in a few clicks:

1. Fork this repo
2. Go to vercel.com/new and import your fork
3. No environment variables needed — just deploy
4. Your server will be at https://your-project.vercel.app/mcp

Dev commands

npm run build:mcp  # Builds the mcp server layer which generates mcpb file
npm run build:ui   # Builds the ChessAgine MCP UI html files
npm run build      # Builds entire project, use for local development
npm run start      # starts the MCP server
npm run debug      # opens MCP inspector to inspect new changes made

License

This project is licensed under the MIT License, the /themes and /protocol are under GPL. See the LICENSE file for details.

Authors

@jalpp