Apache Burr: Build reliable AI agents and applications

I'm still on the fence about agent frameworks, they have their place, and it depends on the nature of the agent: e.g. "Low latency, return a good enough response in 3 seconds, vs. working for 3 hours on a problem."

BUT, if you boil it down, an agent really is context building, making an LLM call, executing requested tool calls, parsing the final model output, returning it to some frontend. There's extensions like memory, async tool calls, etc, but not THAT complicated from a traditional software engineering perspective.

Everyone seems to want to build their agent framework. But if you're tasked with building an agent, I've found it much easier and more maintainable to just build 1:1 code for THAT agent: most of the abstractions you get from an agent framework purely get in the way and obfuscate core agent logic.

You end up being forced to use the abstractions chosen by the agent framework, which sometimes are a mismatch for what you're actually trying to do.

Was this built using https://vorpus.github.io/performativeUI/?

It hits every AI generated landing page trope possible.

Or was it done ironically?

How does this compare to https://strandsagents.com/ ? I'm interested in tools in this space, right now I'm not attached to one, but Bedrock + Serverless on Agent Core feels like the "easy guided path" though I don't like the platform lock-in

I have enjoyed using this framework in my personal and work projects, having a reliable stateful workflow for AI models while getting free observability. I stitched a tool that allows mounting a Burr state machine as an MCP, giving agents a rail to follow, and no matter how complex the state machine gets the MCP tools are constrained to state machine navigation: https://github.com/msradam/theodosia

I am currently working on skills-to-state-machine conversions, since a lot of popular skills out there are already written as phases for an AI model to follow, so it would be great to leverage the explicit functionality of Burr to make that more reliable. Thank you for this amazing project.

A builder pattern and decorators.

Yes, Python has decorators, but they're best used as "filters" that apply to functions or methods. Cache this, serialize the output of this function always, prepare this function to be used as a tool by an agentic harness. Not registration, not flow control. You may disagree but someone has to say it; FastAPI influenced the modern use of decorators far too much in the wrong direction.

Builder patterns are a Rust convention, because Rust has no named keyword arguments. A Python function already exposes a named contract. There is very little reason to ever to sequentially pass configuration parameters in chained method calls. If you need to add state that doesn't exist yet to a constructor or factory, that is not a builder pattern. That is registration. The one place where builder patterns should be tolerated is query builders. They iteratively build on a concept and having the additional "slot" for metadata (method name plus keyword arguments) is genuinely useful. Using methods which accept single parameter instead of keyword arguments is incorrect.

First time I hear about Burr, curious why it was incubated in Apache.

I've been working with jido https://jido.run and would definitely recommend it

I couldn't find an explicit reference for the naming, but for anyone wondering there is a Hamilton example: https://github.com/apache/burr/tree/main/examples/multi-agen...