Burr models applications as state machines where actions transition between states based on conditions. This explicit structure makes agent behavior predictable, debuggable, and reproducible.
Automatic state persistence saves application state at every transition, enabling pause/resume, crash recovery, and human-in-the-loop workflows. Time-travel debugging replays past executions step by step.
The Burr UI provides visual inspection of execution flows, showing state transitions, action inputs/outputs, and decision points. This observability is invaluable for debugging complex agent behaviors.
Integrates with LangChain, OpenAI, and any Python code. Apache 2.0 licensed. Particularly valuable for teams building AI applications that need auditability and reproducibility.