lichess-review-mcp
Read-only MCP server for reviewing completed public Lichess games, providing tools to fetch a game or list recent completed games with strict safety and cost boundaries.
README
lichess-review-mcp
Read-only MCP server for reviewing completed public Lichess games in ChatGPT. It fetches a game or a small recent-game list and refuses ongoing, missing-status, or unknown-status games.
Tools
get_completed_game: accepts an eight-character game ID or exact-hosthttps://lichess.orggame URL.list_recent_completed_games: accepts a public username and returns 5 games by default, at most 10.
Safety and cost boundaries
- Read-only; no OAuth, database, move playing, messaging, challenges, or account management.
- No server-side Stockfish or bulk history imports.
- Fixed Lichess origin, one upstream request at a time, bounded queue, 8-second timeout, manual redirect refusal, 512 KiB response limit, no automatic retries, and at least 60 seconds of cooldown after HTTP 429.
- Small in-memory TTL/LRU caches only.
- Completed-game status allowlist.
created,started, missing, and unknown statuses fail closed. - Localhost-only Host/Origin defaults. A deployment must explicitly set
ALLOWED_HOSTSandALLOWED_ORIGINS. - HTTP backstops default to 120 MCP requests per minute globally and 8 in-flight MCP requests. Excess work is rejected before parsing or dispatch.
- Optional structured request logs contain only a generated request ID, method, route class, response status, and duration. Tool arguments, usernames, game IDs, IP addresses, hosts, origins, and raw paths are not logged.
Local use
Requires Node.js 22 or newer.
npm ci
npm run check
npm test
npm run build
npm run smoke
npm start
Default endpoints:
- Health:
http://127.0.0.1:3000/health - MCP:
http://127.0.0.1:3000/mcp
For a tunnel or deployment, set exact comma-separated values, for example:
ALLOWED_HOSTS=example-tunnel.example \
ALLOWED_ORIGINS=https://chatgpt.com \
HOST=0.0.0.0 PORT=3000 npm start
HTTP hardening can be tuned with positive integers:
HTTP_RATE_LIMIT_REQUESTS=120 \
HTTP_RATE_LIMIT_WINDOW_MS=60000 \
HTTP_MAX_IN_FLIGHT_REQUESTS=8 \
LOG_REQUESTS=1 \
npm start
The built-in request limiter is a process-wide emergency backstop, not a substitute for per-client limits at a trusted reverse proxy or edge. The /health endpoint remains minimal and is excluded from request logs and MCP request limits.
Do not deploy publicly until the integration tests, deployment limits, publisher requirements, and support ownership are separately approved.
Protocol surface
The server implements the stateless JSON-response portion of Streamable HTTP needed for initialization, ping, tools/list, and tools/call. It returns HTTP 202 for notifications and HTTP 405 for unsupported GET/DELETE streaming sessions.
Documentation
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。