Empower AI agents and applications with high-speed deterministic Stockfish position evaluations, FEN game-state explanations, and computer vision board recognition over REST & MCP.
LLMs are non-deterministic pattern matchers prone to hallucinating illegal chess moves or fake checkmates. Chess is strictly deterministic—let Cheetah evaluate positions deterministically while your LLM focuses on strategy.
Your AI agent doesn't need to compile engine binaries, manage local engine runtimes, or set up complex native operating system dependencies. Simply make simple HTTP requests.
Avoid writing complex low-level UCI string parsers and process communication loops. Cheetah converts raw engine evaluation output into clean, structured JSON.
Upload an image file or supply an image URL of a chessboard screenshot. Cheetah automatically crops and detects piece layouts to extract board FEN representations.
Native MCP server integration over Streamable HTTP allowing Claude Desktop, Cursor, or AI agent frameworks to consume engine tools directly out of the box.
Built-in intelligent caching delivers instant sub-millisecond responses on evaluated positions, eliminating redundant calculations.
Evaluates a FEN position using Stockfish. Returns best move (algebraic & SAN), evaluation score (centipawns or mate), and principal variations.
Parses a FEN string using the Fen utility and returns structured game state details (whose turn, castling rights, in_check, game_over, checkmate/stalemate status, winner).
Recognizes chessboard layout from an uploaded image file or URL. Accepts an optional flip=1 parameter to rotate the output 180°.
Model Context Protocol (MCP) Streamable HTTP endpoint for AI tool calling. Provides evaluate_position and explain_position tools.