I built Baikon, a logic-first DSL for creating AI agents using clean, readable flows.

Instead of wiring together Python classes or YAML configs, you just write:

flow assistant: when user says "hello" -> call greet_user

function greet_user: say "Hi {user_name}, I’m your assistant."

1. Human-readable syntax 2. Built-in variable and state support 3. Async-capable CLI engine 4. GPT/Claude prompt-friendly 5. Lightweight and open-source

GitHub: https://github.com/baikondev/baikon Website: https://baikon.dev Discord: https://discord.gg/xN4nXt5Z X: https://twitter.com/BaikonDev producthunt.com/@baikon_dev

Would love feedback, contributions, or ideas from other AI builders.

I built Baikon because I was frustrated with how complex AI agent development has become. Instead of scattered Python files and complex frameworks, you can write:

flow assistant: when user says "hello" -> call greet_user

function greet_user: say "Hi there! I'm your assistant."

That's a complete working AI agent. The syntax is human-readable but executable.

It includes an async engine, middleware system, variable management, and API integration - all the enterprise features you need, but with logic that anyone can understand.

I'd love feedback from the HN community. What would you build if AI agent logic was this simple?