apache/burr

What it solves

Apache Burr provides a structured way to build stateful applications, such as AI agents and chatbots, by modeling them as state machines. It solves the complexity of managing application state, tracking decision flows, and debugging agentic workflows, which often become difficult to maintain in standard script-based implementations.

How it works

Developers define their application logic as a series of Python functions (actions) and transitions between them, creating a graph-like structure. The framework handles the state transitions and persistence. It includes a built-in telemetry UI that allows developers to visualize the execution flow, inspect state changes, and debug the application in real-time as it runs.

Who it’s for

It is designed for developers building complex, stateful applications—particularly those involving LLMs—who need a framework-agnostic way to orchestrate logic, manage memory, and gain visibility into their agent's decision-making process.

Highlights

• Explicit state machine modeling for predictable, manageable application workflows.
• Real-time telemetry UI for tracing and debugging execution state.
• Framework-agnostic design that integrates with any LLM provider or Python library.
• Pluggable persistence layers for saving and loading application state.
• Supports both LLM-based applications and general-purpose stateful simulations.