Overview
A clean starting point for building agents in Python, so you can focus on your use case instead of the plumbing.
It works with both the OpenAI and Anthropic SDKs and is MIT-licensed, so you can use it in your own projects.
Key highlights
What's inside
Project structure
Where agents, tools and prompts live.
Tools
Defining tools with validated inputs.
Memory
Conversation memory and a simple long-term store.
Output parsing
Getting reliable structured data back.
Multi-agent
A supervisor with two specialist agents.
Safety
Rate limits, cost caps and approval steps.
Tools & technologies
Is this for you?
Who it's for
Developers starting an agent project.
Before you start
Python and an API key for OpenAI or Anthropic.
