Chess
Max Health Inc.

Play chess against your AI

ChatGPT, Claude, or the built-in Stockfish engine: pick your opponent and play in seconds. Then let the engine break down any game move by move: accuracy, blunders, and the turning points.

by Max Health Inc.

97 players have already joined the multiplayer lobby

Add to ChatGPT →Try it out ↗

Try it out opens AIHR, the Max Health record this project is a playground for.

Watch demo →
01 / Getting Started

Connect in seconds

Add to ChatGPT in one click, or add the connector below in any other assistant.

01
Open Claude, ChatGPT, or another AI assistant that supports connectors
02
Add a connector pointing at https://chess.maxhealth.tech/mcp
03
Ask your AI to play chess with you or enter the multiplayer lobby
Play chess with me
Here's your board. You're white, make your move!
02 / Motivation

Why chess, why healthcare

Chess has long served as a benchmark for artificial intelligence. Dedicated engines like Stockfish and AlphaZero play at superhuman levels using search trees, evaluation functions, and reinforcement learning. They prove that machines can master complex decision-making when the boundaries are well defined.

Healthcare shares that aspiration: use AI to make better decisions, faster. Train systems to recognize patterns, weigh options, and recommend the action that leads to a better outcome.

So this project puts both kinds of AI on one board. You can play the bundled Stockfish engine, exact and calculating and reliable, or play a large language model like ChatGPT or Claude, which approximates reasoning by pattern-matching on its training data instead of calculating lines. We keep the model legal by only offering it valid moves and checking its choice server-side. Even so, it will miscalculate positions, miss tactics, and confidently propose losing plans.

Those are the very failure modes we see when LLMs are applied to healthcare: overconfident recommendations, missed context, and blind spots in the edge cases. The difference here is that the same engine grades every move, so each mistake is visible immediately rather than discovered too late.

That makes this a live playground for the question healthcare actually faces: where does general-purpose AI fall short, and how do we measure it against the truth? The engine is the ground truth; the language model is the system under test. It is exactly the kind of evaluation we need to build AI you can trust with far more than a chess game.

03 / Questions

Frequently Asked Questions

What do I need to play?
An AI assistant that supports connectors, such as Claude or ChatGPT. Add https://chess.maxhealth.tech/mcp as a connector and ask it to play chess. There is nothing to download: the board appears inside the conversation, and you move pieces by clicking it.
Do I need an account?
No. Connecting offers "Continue as guest", and a guest can play solo, enter the lobby and analyse games. Signing in with Max Health only adds what needs an identity to mean anything: a claimed handle, a ranked ELO rating, and a match history that survives the session.
Who am I actually playing against?
Whoever you pick. The built-in Stockfish engine plays at six levels, from beginner to genuinely strong. The host model, the ChatGPT or Claude you are already talking to, can play the moves itself instead. Or enter the lobby and play another person in real time.
How strong is the engine?
Strong enough that it is deliberately dialled down. Six levels run from a gentle opponent to one that punishes every mistake. The same engine grades your moves afterwards, and that is what makes the analysis worth reading: the evaluation comes from a real engine search, not from a language model guessing.
Can it analyse games I played somewhere else?
Yes. Give it any public chess.com or Lichess handle and it reads their recent games and reports the mistakes that keep coming back. A single game can be reviewed move by move: accuracy, blunders, the better move you missed, and the turning points where the result was decided.
Does it cost anything?
No. Playing and analysis are free. There is no paid tier and nothing is metered.
What do you store?
Live games and lobby presence are temporary and are cleared automatically once a game is abandoned or a player goes quiet. If you sign in, your handle, your rating and your finished games are kept so the rating means something. The privacy policy has the specifics.
Why is a health company building a chess app?
Because the decision has the same shape. In chess, every move is judged by whether it keeps the king. In medicine, by whether it keeps the heartbeat. Chess is one of the few places where a machine can be held to that judgement on every single move: the engine is the truth, and the language model is the one under test. The mistakes it makes here, confident and unaware of being wrong, are exactly the mistakes that matter when the same kind of model is pointed at a health record.
EnglishDeutschItalianoFrançaisEspañol