Master Apache Burr with our step-by-step tutorial, detailed feature walkthrough, and expert tips.
Explore the key features that make Apache Burr powerful for coding agents workflows.
Define applications as explicit state machines with decorator-based actions and conditional transitions, then inspect and replay executions through the built-in Burr UI's trace viewer and state inspector.
Works seamlessly with any LLM provider (OpenAI, Anthropic, local models) and Python library, with no vendor lock-in or required abstraction layers.
Every installation includes a local web UI for step-by-step execution traces, state inspection, and time-travel debugging — no external services or accounts required.
Pluggable persistence backends support in-memory, SQLite, PostgreSQL, Redis, and custom stores for checkpointing, recovery, and long-running workflow state.
Deploy applications as web services with built-in FastAPI support, enabling straightforward scaling and integration into existing service architectures.
Currently incubating at the ASF, benefiting from its proven governance model with vendor-neutral oversight, transparent development, and community-driven roadmap.
Deep introspection capabilities allow examination of state at every step, enabling reproducible debugging and comprehensive audit trails for compliance.
Seamlessly manages state containing text, images, embeddings, and structured data across actions and transitions in complex AI pipelines.
While basic understanding helps, Burr's state machine model is intentionally simple. Actions define inputs and outputs, and transitions specify which action runs next. The decorator-based API makes it feel like writing standard Python functions with clear control flow.
Burr is completely framework-agnostic. Actions are standard Python functions, so you can call OpenAI, Anthropic, local models via Ollama, or any other provider. There is no built-in LLM abstraction layer that forces you into a specific integration.
Burr's telemetry UI is built-in and free, providing step-by-step execution traces, state inspection, and time-travel debugging out of the box. LangSmith is a separate paid service starting at $39 per seat per month. Burr's approach requires no external accounts or API keys for local debugging.
Yes. Burr includes FastAPI integration, persistent state backends, and robust error handling suitable for production. Its Apache Software Foundation incubation status signals community commitment to long-term maintenance and governance. Note that the project is still in ASF incubation, so users should evaluate maturity for their specific requirements.
Burr's overhead is minimal since it primarily orchestrates function calls and manages state transitions. The actual computational work happens in your actions (LLM calls, data processing), and Burr adds negligible latency to the orchestration layer.
Migration involves restructuring chain logic into actions and transitions. Since Burr actions are plain Python functions, existing LangChain tool integrations can often be wrapped directly. The main effort is in redesigning the flow as an explicit state machine.
The open-source version includes community support via Discord and GitHub. Burr Cloud (currently in beta) is planned to offer hosted observability and team features. Beta access is currently free; post-GA pricing has not been publicly announced but is expected to follow industry-standard per-seat or usage-based models. The Apache Software Foundation governance model ensures the project's long-term continuity regardless of commercial offerings.
Yes. Burr applications can run concurrently with isolated state, making it straightforward to orchestrate multiple agents or parallel workflows within a single service.
Now that you know how to use Apache Burr, it's time to put this knowledge into practice.
Sign up and follow the tutorial steps
Check pros, cons, and user feedback
See how it stacks against alternatives
Follow our tutorial and master this powerful coding agents tool in minutes.
Tutorial updated March 2026