While building with AI agents, I kept running into the same problem: testing even tiny changes to a prompt or tool required either running my entire app or writing throwaway scripts. Both were slow and constantly broke my flow.
I’ve always loved how Storybook lets you work on UI components in isolation, so I tried applying the same idea to agents.
Agent Playbook is an open-source, Storybook-like playground for developing and debugging AI agents.
It automatically discovers your agents, spins up a local web UI, and gives you an interactive environment to test them. You can chat with an agent, inspect its reasoning steps, and iterate quickly with hot-reloading.
GitHub: https://github.com/orlevii/agent-playbook
Right now it supports agents built with pydantic-ai, which I’ve been using heavily and found easy to work with. I’m planning to expand support based on what people find useful, so feedback is especially welcome.
This is the first public release, and I’d love feedback on the core idea, the developer experience, and any feature ideas that would make it more useful. Happy to answer questions!
Thanks for taking a look.
Comments URL: https://news.ycombinator.com/item?id=45925723
Points: 2
# Comments: 0
Source: github.com
