_action
ChatGPTInternal tool called by the chess UI. Do not call directly.
Entertainment
Overview
Chess by Max Health lets people play and study chess inside ChatGPT. Players can start a single-player game against the built-in Stockfish engine at a chosen difficulty (Beginner to Maximum) or against the conversation model, with selectable color and time controls, and can join a live multiplayer lobby to play other people (with optional ranked ELO and match history for signed-in Max Health users). The same Stockfish engine analyzes games: per-move accuracy, blunders and inaccuracies, the turning points, and the time spent on each move, shown on an interactive board the user can step through. Games to analyze can come from the current session, a signed-in user's stored matches, or any public chess.com or Lichess username fetched live. Single positions can also be evaluated for the best move and main line, and key positions can be illustrated on a board with the recommended move and an evaluation bar.
Tools
_action
ChatGPTInternal tool called by the chess UI. Do not call directly.
analyze_game
ChatGPTAnalyze one chess game with the Stockfish engine: per-move evaluation, best & worst moves (blunders with the better move), accuracy, the hottest (most decisive) positions, and — when the game has clocks — time used per move (analysis.timeInsights). Sources: a Max Health game by gameId (with username to pick your side), a bare SAN
moves list (with color:"white"|"black"), or source:"chess.com" + username to analyze that player's most recent public game.analyze_game
ChatGPTAnalyze one chess game with the Stockfish engine: per-move evaluation, best & worst moves (blunders with the better move), accuracy, the hottest (most decisive) positions, and — when the game has clocks — time used per move (analysis.timeInsights). Sources: a Max Health game by gameId (with username to pick your side), a bare SAN
moves list (with color:"white"|"black"), or source:"chess.com" + username to analyze that player's most recent public game.analyze_history
ChatGPTFetches and analyzes a player's recent games. Pass
username (a chess.com or Lichess handle, e.g. "MagnusCarlsen", with optional count up to 20) plus source ("chess.com" default, or "lichess") to analyze their public games straight from that site's API — no account linking or PGN needed. Omit username to analyze the signed-in Max Health user's own stored games instead (requires sign-in). Returns recurring patterns plus, per game, its gameId, the player's userColor and worst moves (each with fenBefore, uci, bestUci), time usage, and that game's full moves list.analyze_history
ChatGPTFetches and analyzes a player's recent games. Pass
username (a chess.com or Lichess handle, e.g. "MagnusCarlsen", with optional count up to 20) plus source ("chess.com" default, or "lichess") to analyze their public games straight from that site's API — no account linking or PGN needed. Omit username to analyze the signed-in Max Health user's own stored games instead (requires sign-in). Returns recurring patterns plus, per game, its gameId, the player's userColor and worst moves (each with fenBefore, uci, bestUci), time usage, and that game's full moves list.enter_lobby
ChatGPTJoin the chess lobby with a username to play against other players.
enter_lobby
ChatGPTJoin the chess lobby with a username to play against other players.
evaluate_position
ChatGPTEvaluate ANY chess position (given a FEN) with the Stockfish engine: returns the score from White's perspective (whiteCp + a human label like "+1.4" or "M3"), the best move (bestUci + bestSan) and the principal variation (pvSan). Use this to answer "what should I play here?", "who is better?" or "is this winning?" for a single position. To SHOW the recommendation, follow up with illustrate_position (fen, arrow=bestUci, evalCp=whiteCp).
evaluate_position
ChatGPTEvaluate ANY chess position (given a FEN) with the Stockfish engine: returns the score from White's perspective (whiteCp + a human label like "+1.4" or "M3"), the best move (bestUci + bestSan) and the principal variation (pvSan). Use this to answer "what should I play here?", "who is better?" or "is this winning?" for a single position. To SHOW the recommendation, follow up with illustrate_position (fen, arrow=bestUci, evalCp=whiteCp).
illustrate_position
ChatGPTIllustrate ONE chess position while you EXPLAIN analysis: shows the board with an optional highlighted move, a green arrow for the better move, an eval bar, and a caption. Do NOT use it to start or show a live game (play_computer / enter_lobby already render the interactive board) — only for narrating analysis of past positions, one call per position. After analyze_game, use each move's fenBefore + uci (highlight) + bestUci (arrow).
illustrate_position
ChatGPTIllustrate ONE chess position while you EXPLAIN analysis: shows the board with an optional highlighted move, a green arrow for the better move, an eval bar, and a caption. Do NOT use it to start or show a live game (play_computer / enter_lobby already render the interactive board) — only for narrating analysis of past positions, one call per position. After analyze_game, use each move's fenBefore + uci (highlight) + bestUci (arrow).
llm_move
ChatGPTPlay a chess move as the computer opponent. Called by the LLM when sampling is unavailable. Pick the best strategic move from the legal moves provided in the prompt.
llm_move
ChatGPTPlay a chess move as the computer opponent. Called by the LLM when sampling is unavailable. Pick the best strategic move from the legal moves provided in the prompt.
play_computer
ChatGPTStart a chess game against the AI. The LLM picks its moves via MCP sampling.
play_computer
ChatGPTStart a single-player chess game. opponent="host" lets you (the model) play; opponent="engine" plays the built-in Stockfish engine at the chosen difficulty level.
Capabilities
WritesInteractive
Example Prompts
Click any prompt to copy it.
