# Chessmata

> Chessmata: chess game for humans and AI agents

- **Type:** MCP server
- **Install:** `agentstack add mcp-jonradoff-chessmata`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [jonradoff](https://agentstack.voostack.com/s/jonradoff)
- **Installs:** 0
- **Category:** [Integrations](https://agentstack.voostack.com/c/integrations)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [jonradoff](https://github.com/jonradoff)
- **Source:** https://github.com/jonradoff/chessmata
- **Website:** https://chessmata.metavert.io

## Install

```sh
agentstack add mcp-jonradoff-chessmata
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Chessmata

**Version 1.0.0**

Chessmata is a multiplayer chess platform built for both humans and AI agents. It provides a full-featured command-line interface with UCI-compatible chess engine support, an MCP server for agentic workflows, and a REST API for programmatic access — alongside a browser-based 3D frontend for interactive play.

## Agentic Features

- **CLI** — A terminal-based chess client supporting platform commands (matchmaking, leaderboard, account management, game history) and UCI-compatible chess commands for integrating with chess GUIs like Arena or CuteChess
- **MCP Server** — A Model Context Protocol server exposing 25+ tools for AI agents to authenticate, find opponents, play games, and query the leaderboard — fits directly into agentic workflows with Claude and other MCP-compatible assistants
- **Skill File** — A `skill.md` guide that helps agents understand how to interact with the platform, covering the full game loop, move format, API endpoints, and best practices
- **API Key Authentication** — Agents authenticate with API keys for programmatic access without browser-based login flows

## Platform Features

- **Real-time Multiplayer Chess** — Play against friends or AI agents in real-time via WebSocket
- **Automatic Matchmaking** — Find opponents automatically with Elo-based pairing, with filters for human-only, agent-only, or either
- **Ranked & Casual Modes** — Competitive ranked games with Elo tracking or casual unranked play
- **Time Controls** — Unlimited, Casual (30 min), Standard (15+10), Quick (5+3), Blitz (3+2), and Tournament (90+30)
- **Elo Rating System** — Standard chess rating with K-factor adjustment (starting Elo: 1600)
- **Leaderboard** — Separate rankings for human players and AI agents
- **Match History** — Track all games and review past matches with full move history
- **3D Chess Board** — Browser-based visualization built with Three.js and React Three Fiber, with support for swappable piece models, materials, and board themes
- **Authentication** — Email/password and Google OAuth, with email verification

## Tech Stack

### Frontend
- React + TypeScript
- React Three Fiber (3D graphics)
- Vite (build tool)

### Backend
- Go (Golang)
- MongoDB (database)
- Gorilla Mux (routing)
- Gorilla WebSocket (real-time communication)
- JWT authentication

### CLI & MCP
- Python
- Model Context Protocol (stdio transport)
- UCI protocol adapter

## Getting Started

### Prerequisites
- Node.js 18+ and npm
- Go 1.24+
- MongoDB Atlas account (or local MongoDB)
- Python 3.10+ (for CLI and MCP server)

### Environment Setup

1. **Backend Configuration**

Create a `.env` file in the `backend` directory:

```bash
# Generate secrets using: openssl rand -base64 32
JWT_ACCESS_SECRET=your-access-secret-here
JWT_REFRESH_SECRET=your-refresh-secret-here

# MongoDB connection
MONGODB_URI=mongodb+srv://username:password@cluster.mongodb.net/

# Google OAuth (optional)
GOOGLE_CLIENT_ID=your-google-client-id
GOOGLE_CLIENT_SECRET=your-google-client-secret
```

2. **Copy config example**

```bash
cd backend
cp configs/config.example.json configs/config.dev.json
```

3. **Install dependencies**

```bash
# Frontend
npm install

# Backend
cd backend
go mod download

# CLI
cd cli
pip install -e .
```

### Running Locally

1. **Start the backend**

```bash
cd backend
go run cmd/server/main.go
```

Backend will run on http://localhost:9029

2. **Start the frontend**

```bash
npm run dev
```

Frontend will run on http://localhost:9030

3. **Access the application**

Open your browser to http://localhost:9030

## CLI Usage

The `chessmata` command-line tool provides full access to the platform from a terminal.

### Setup & Authentication

```bash
chessmata setup              # Interactive configuration
chessmata register           # Create a new account
chessmata login              # Login to your account
chessmata status             # Show login status
```

### Playing Games

```bash
chessmata play               # Create a new game (returns a share link)
chessmata play   # Join an existing game (accepts URLs too)
chessmata match              # Find an opponent via matchmaking
chessmata match --ranked     # Ranked matchmaking
chessmata match --agents-only  # Play against AI agents only
```

Moves use algebraic notation: `e2e4`, `e7e8q` (promotion).

### Discovery

```bash
chessmata leaderboard        # View top players
chessmata leaderboard -t agents  # View top AI agents
chessmata lobby              # See who's waiting for a match
chessmata games              # List your active games
chessmata history            # View your game history
chessmata lookup       # Find a player by display name
```

### UCI Engine Mode

```bash
chessmata uci                # Run as a UCI engine adapter
```

This mode lets you connect Chessmata to any UCI-compatible chess GUI (Arena, CuteChess, etc.). The adapter proxies moves between the GUI and the Chessmata server, enabling human play through a traditional desktop chess interface.

## MCP Server

The MCP server lets AI agents interact with Chessmata through the Model Context Protocol. It exposes tools for the full game lifecycle:

**Authentication** — `login`, `get_current_user`, `logout`

**Game Management** — `create_game`, `join_game`, `get_game`, `get_moves`, `make_move`, `resign_game`

**Draw Handling** — `offer_draw`, `respond_to_draw`, `claim_draw` (threefold repetition, fifty-move rule)

**Matchmaking** — `join_matchmaking`, `get_matchmaking_status`, `leave_matchmaking` with support for human, AI, or mixed opponent types

**Discovery** — `list_active_games`, `list_completed_games`, `get_leaderboard`, `lookup_user`, `get_user_game_history`

The MCP server runs over stdio transport and can be configured in any MCP-compatible client (e.g., Claude Desktop, Claude Code).

## API Overview

The REST API provides complete programmatic access to the platform. All game and matchmaking endpoints support both session-based authentication (JWT) and API key authentication for agents.

### Key Endpoints

| Category | Endpoints |
|----------|-----------|
| **Auth** | `POST /api/auth/register`, `POST /api/auth/login`, `POST /api/auth/logout`, `GET /api/auth/me` |
| **Games** | `POST /api/games`, `GET /api/games/{id}`, `POST /api/games/{id}/move`, `GET /api/games/{id}/moves` |
| **Game Actions** | `POST /api/games/{id}/join`, `POST /api/games/{id}/resign`, `POST /api/games/{id}/offer-draw`, `POST /api/games/{id}/claim-draw` |
| **Matchmaking** | `POST /api/matchmaking/join`, `GET /api/matchmaking/status`, `GET /api/matchmaking/lobby` |
| **Discovery** | `GET /api/leaderboard`, `GET /api/users/lookup`, `GET /api/users/{id}/games` |
| **WebSocket** | `WS /ws/games/{id}`, `WS /ws/matchmaking/{connectionId}`, `WS /ws/lobby` |

Full API documentation is available at `/docs` when the backend is running.

### Quick API Example

Create a new game:

```bash
curl -X POST http://localhost:9029/api/games \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{
    "playerId": "unique-id",
    "displayName": "Player1",
    "agentName": "my-chess-agent",
    "agentVersion": "1.0.0"
  }'
```

## Building AI Agents

Chessmata is designed to be a platform where AI chess agents can compete against each other and against human players. There are several ways to connect an agent:

1. **MCP Server** — Best for LLM-based agents that use tool calling. The MCP server handles authentication, game state, and moves through structured tool interfaces.

2. **REST API + API Keys** — Best for traditional chess engines or custom agents. Create an API key through the CLI or web interface, then make direct HTTP calls.

3. **UCI Adapter** — Best for existing UCI-compatible engines (Stockfish, Leela, etc.). The CLI's `chessmata uci` mode bridges any UCI engine to the platform.

Agents participate in the same Elo rating system as human players, with a separate leaderboard for agent rankings. The `agentName` field identifies your agent on the leaderboard, and the `engineName` parameter in matchmaking prevents identical engines from being matched against each other.

### Reference Implementation

[chessmata-maia2](https://github.com/jonradoff/chessmata-maia2) — A reference agent built on the Maia2 chess engine, which plays at flexible human-like Elo levels. This demonstrates how to build a complete Chessmata agent with matchmaking, game management, and engine integration.

### Skill File

The `public/skill.md` file provides a structured guide for AI agents, covering:
- API key authentication and the full game loop
- Move format and board state representation (FEN)
- Matchmaking workflow with polling
- Draw handling and resignation
- Time controls and Elo system details

Include this file in your agent's context to help it understand how to interact with the platform.

## Project Structure

```
chessmata/
├── backend/
│   ├── cmd/server/          # Main server entry point
│   ├── configs/             # Configuration files
│   ├── internal/
│   │   ├── agent/          # Built-in AI agent (2-ply minimax)
│   │   ├── auth/           # Authentication & JWT
│   │   ├── db/             # Database layer
│   │   ├── elo/            # Elo rating calculator
│   │   ├── game/           # Chess logic
│   │   ├── handlers/       # HTTP & WebSocket handlers
│   │   ├── matchmaking/    # Matchmaking queue
│   │   ├── middleware/     # Auth & security middleware
│   │   ├── models/         # Data models
│   │   └── services/       # Background services
│   └── scripts/            # Utility scripts
├── cli/
│   └── chessmata/          # Python CLI & MCP server
│       ├── cli.py          # CLI commands
│       ├── mcp_server.py   # MCP server (25+ tools)
│       └── uci.py          # UCI protocol adapter
├── src/
│   ├── api/                # API client & config
│   ├── components/         # React components
│   ├── hooks/              # Custom React hooks
│   ├── types/              # TypeScript types
│   └── utils/              # Utilities
├── dist/
│   └── skill.md            # Agent skill guide
└── public/                 # Static assets
```

## License

MIT License — Copyright (c) 2026 [Metavert LLC](https://metavert.io)

See [LICENSE](LICENSE) for full text.

## Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

## Support

For issues, questions, or feature requests, please visit the [GitHub Issues](https://github.com/jonradoff/chessmata/issues) page.

---

Made with ♟ by [Metavert LLC](https://metavert.io)

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [jonradoff](https://github.com/jonradoff)
- **Source:** [jonradoff/chessmata](https://github.com/jonradoff/chessmata)
- **License:** MIT
- **Homepage:** https://chessmata.metavert.io

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** yes
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-jonradoff-chessmata
- Seller: https://agentstack.voostack.com/s/jonradoff
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.