Sunfish is a Python chess engine that implements a complete UCI-compatible chess AI in minimal code.
The engine solves the problem of creating a functional chess AI by combining alpha-beta search with a simple evaluation function, achieving competitive play despite its extreme brevity. The approach uses negamax search with quiescence to handle tactical positions, allowing the engine to play reasonably strong chess while remaining simple enough to understand and modify. The project includes both a standard version and an NNUE-based variant that uses a small neural network for evaluation.
Sunfish suits developers who want to experiment with chess engine design, test search algorithms, or learn how chess engines work without navigating a large codebase. It works well as a foundation for research into parallel search, evaluation function tuning, or deep learning approaches to chess. The tool integrates with standard chess GUIs and tournament software through its UCI protocol support, and can be packed into a tiny self-extracting executable. The README explicitly positions it as a learning platform and experimental base rather than a production engine competing with established alternatives.
The project maintains active engagement with its codebase through continuous integration testing of UCI protocol compliance and tournament-style matches. Development includes documented testing methodology and support for multiple deployment scenarios, from terminal play to GUI integration to compressed executable formats. The maintainers have extended the core concept with an NNUE variant available for online play, demonstrating ongoing evolution of the original design.
