AutoGen

AutoGen is an open-source framework from Microsoft Research. It builds apps where teams of AI agents work and think together. Instead of one chatbot, you get a team of agents that solve problems together. Its AgentChat API handles the flow between agents and tools. A no-code tool called AutoGen Studio lets you test agents in a browser. The code is MIT-licensed and free for anyone to use. Note that the project is in maintenance mode. Microsoft points new users to its Agent Framework.

Key Features

1. AgentChat conversation API

Agents exchange messages in structured chats to split up work. You define each agent's role, and the framework routes the conversation.

2. AutoGen Studio GUI

Build and test agents in a no-code web interface. Install it with pip and open it in your browser to prototype fast.

3. Works with many AI models

Use OpenAI models or others through extension packages. You are not locked into a single model provider.

4. Tool and web-browsing support

Agents can call functions, run code, and browse the web through MCP servers. The Playwright MCP adds full browser control.

5. Human-in-the-loop chats

You can join any agent conversation to approve steps or give guidance. This keeps people in charge of important decisions.

FAQ

Is AutoGen still being developed?

No new features are planned. Microsoft keeps it in maintenance mode and points new projects to the Microsoft Agent Framework.

What do I need to run AutoGen?

Python 3.10 or newer and an API key from an LLM provider like OpenAI. Install it with pip install -U "autogen-agentchat" "autogen-ext[openai]".

How is AutoGen different from a single chatbot?

A chatbot gives one answer at a time. AutoGen runs a team of agents that debate and divide tasks before answering.

What license does AutoGen use?

The Python packages are MIT-licensed, so you can use them in commercial projects. The docs site itself uses a Creative Commons license.

Conclusion

AutoGen still works well for experiments with teams of AI agents, backed by Microsoft Research. The AgentChat API and Studio GUI make multi-agent ideas easy to test in hours. It is in maintenance mode, so new projects should compare Microsoft's Agent Framework first. For learning and prototypes, it still teaches the core patterns of agent teamwork well.