AutoGen (AG2)
A multi-agent conversation framework where agents collaborate via message-passing, with group chat and code execution.
AutoGen pioneered the conversational multi-agent pattern: agents (and humans) collaborate by passing messages, including group chats and a code-executing agent. It originated at Microsoft Research; AG2 is the community-driven fork that continues that lineage. Both are open source.
AutoGen is an open-source framework that models multi-agent systems as conversations: specialized agents — and optionally a human — exchange messages to solve a task together, including multi-agent group chats and a built-in code-executing agent that can write and run code in a loop. It helped popularize the conversational multi-agent pattern that many later frameworks built on.
It is aimed at developers and researchers prototyping collaborative or self-correcting agent systems. A note on naming: AutoGen originated at Microsoft Research; AG2 is the community-driven fork (formerly AutoGen) that carries the project forward, so you'll see both names in the ecosystem.
Highlights
- Conversable agents — agents communicate by passing messages, composing into multi-agent solutions.
- Group chat — orchestrate several agents (and a human) in a shared conversation with a manager directing turns.
- Code execution — a built-in executor agent writes and runs code, enabling generate-run-debug loops.
- Human-in-the-loop — insert a human agent at any point in the conversation.
- Model-flexible — works across LLM providers.
In an AI-assisted workflow
A common pattern is an assistant agent that proposes code and a user-proxy/executor agent that runs it and feeds back results, iterating until tests pass — collaboration as a conversation rather than a hardcoded pipeline.
NOTE
Check which distribution you're adopting — Microsoft's autogen or the community ag2 fork — since APIs and momentum can differ. Both are open source under permissive licenses.
Good to know
AutoGen/AG2 is open source and free; you bring your own model provider. Its conversational model is flexible and great for prototyping, but for production you may want the explicit control of LangGraph or the structured roles of CrewAI — see the framework comparison.
Related
- LangGraphA low-level library for building stateful, controllable agents as graphs, with checkpointing and human-in-the-loop.
- CrewAIA Python framework for orchestrating role-playing AI agents as collaborating 'crews', plus event-driven flows.
- OpenAI Agents SDKOpenAI's lightweight, open-source framework for agents — handoffs, guardrails, sessions, and built-in tracing.
- Which Agent Framework in 2026? LangGraph vs CrewAI vs AutoGen vs OpenAI Agents SDK vs Claude Agent SDKA decision guide to the major AI agent frameworks — control vs. abstraction, multi-agent models, state and durability, and which fits your project.