Lichess MCP: connecting Claude to Lichess
A Lichess MCP is a server that lets an AI assistant such as Claude call the Lichess API. The best-known one, karayaman/lichess-mcp, is open source and exposes about 190 tools: your account, game exports, puzzles, the opening explorer, Lichess's cloud evaluations and, if you enable them, playing games. It runs on your own computer with a Lichess token.
MCP, the Model Context Protocol, is the open standard assistants use to call outside tools. A "server" is the program that offers those tools.
Lichess MCP servers at a glance
Checked on October 25, 2026, from each project's README. Stars and dates change.
| Project | What it gives Claude | How you run it | Needs |
|---|---|---|---|
| karayaman/lichess-mcp | About 190 tools in toolsets; about 70 on by default (account, users, games, puzzles, opening explorer, tablebase, analysis). Playing games and tournaments are opt-in | Locally, installed with npm (or Smithery) and added to Claude Desktop's config | A Lichess API token; MIT license |
| Smaller community repos | Various wrappers of the same API, including a Python port of karayaman's server | Locally | Mostly undocumented, no stars yet |
| Chess Coach (ours) | Your finished Lichess and Chess.com games, Stockfish reviews, an interactive board | Remote: paste a URL, sign in | No Lichess token or password |
karayaman/lichess-mcp is the established project here: around 20 stars and recent commits in October 2026. Its README is careful about security. It recommends keeping the token in your system's secret store and ticking only the scopes you need, because some tools act on your account and the model reads text other Lichess users wrote.
It also includes interactive tools for Claude Desktop that show a real board in the chat: a puzzle solver, a game viewer and an opening tree. They use MCP Apps, the same extension Chess Coach's board uses.
Lichess AI analysis through MCP: what the API actually returns
People often install a Lichess MCP hoping for "Lichess AI analysis" of their games. Two of the API's analysis features are worth understanding first, because both come with conditions.
Cloud evaluation. The Lichess documentation describes it as the cached evaluation of a position, "if available". The database holds about 320 million positions, and opening positions have the best chance of being there. A position from the middle of your own game, after a few unusual moves, often won't be.
Game export with evals. When you export a game, Lichess can include engine evaluations and its accuracy figures "when available", which means when the game was already analysed on Lichess. If you never requested analysis, the export is just the moves.
So a Lichess MCP gives Claude excellent data, but it doesn't run an engine on demand. Claude fills the gaps with its own judgment. A language model isn't a chess engine and can misread positions, as we explain in Can Claude analyze chess games?. To check any position you can add a separate Stockfish MCP server, or use Lichess's own free analysis board on the site.
Lichess MCP server vs a review connector
They solve different problems.
- A full Lichess MCP server is for doing things on Lichess from a chat: checking profiles, browsing puzzles, managing challenges or teams. It is free and flexible, and you run and secure it yourself.
- A review connector is for one job: understanding your finished games. It needs an engine, a board and explanations, and it shouldn't need write access to your account.
How Chess Coach handles Lichess games
Chess Coach is a remote MCP server for Claude and ChatGPT. In Claude, connect Chess Coach from its directory listing (setup steps), sign in with your email or Google, and tell it your Lichess username.
list_gameslists your finished Lichess games and can search by month, opponent, result, color, time control or opening.review_gameruns Stockfish on our server on every move, not a cache, so it works on unusual positions too.- The interactive board shows the key moments ①②③, the evaluation graph and arrows (red for your move, green for the better one). Practice mistakes lets you replay the positions where you went wrong.
It reads public games without credentials and never helps with a game in progress. The Free plan includes one in-depth review and 30 position checks a day; Plus is $3.99/month and Pro $7.99/month. Chess Coach is available in Claude’s directory and OpenAI’s plugin directory. Open its listing, install it and sign in with your Chess Coach account. More on what a review contains: AI chess analysis. If you also play on Chess.com, see Chess.com MCP servers, and for Lichess's free review compared with Chess.com's, our comparison.
Chess Coach is not affiliated with Lichess.
Frequently asked questions
Is there an official Lichess MCP server?
Not that we could find as of October 2026. The best-known one, karayaman/lichess-mcp, is an open-source community project under the MIT license that wraps the public Lichess API.
Do I need a Lichess token?
For karayaman/lichess-mcp, yes: you create a personal API token on Lichess and should tick only the scopes you need. Reading public, finished games doesn't need your Lichess password, and Chess Coach doesn't ask for any Lichess token or password.
Does the Lichess cloud evaluation analyze any position?
No. It returns a cached evaluation only if the position is already in Lichess's database of about 320 million positions. Opening positions are often there; unusual middlegame positions from your own games often aren't.
Can I get AI analysis of my Lichess games in Claude?
Yes. With Chess Coach connected, give Claude your Lichess username or a game link: Stockfish reviews the finished game on our server and Claude explains the key moments on an interactive board.