Open Source
Other Multi-AgentMicrosoftPythonCode ExecutionResearch
AutoGen
Research-backed multi-agent framework from Microsoft. Agents converse with each other and humans to solve complex tasks via code execution, tool calls, and planning.
Install
$pip install pyautogen About
AutoGen introduces the concept of conversable agents — LLM-powered agents that can talk to each other, execute code, and call tools in a structured conversation loop. A typical workflow is an AssistantAgent paired with a UserProxyAgent that runs code and returns results.
AG2 (AutoGen 0.4+) is a ground-up async rewrite with an event-driven message system, better support for long-running workflows, and a cleaner API. It is the recommended version for new projects.
Getting Started
pip install pyautogen
import autogen
assistant = autogen.AssistantAgent(
name="assistant",
llm_config={"model": "claude-sonnet-4-5", "api_key": "..."}
)
user_proxy = autogen.UserProxyAgent(name="user_proxy", human_input_mode="NEVER")
user_proxy.initiate_chat(assistant, message="Plot a sine wave and save it as plot.png")Features
Conversational agents with customisable roles and termination conditions
Code execution in Docker or local sandbox with output routing
Human-in-the-loop — any agent can be a proxy for human input
Group chat with speaker selection (auto, round-robin, or custom)
Tool calling with automatic function registration
AG2 (formerly AutoGen 0.4) — fully async event-driven rewrite
AutoGen Studio — no-code GUI for building agent workflows
Active Microsoft Research backing with NeurIPS / ICLR papers
References & Links
Project Info
Language Python
License MIT
Stars 43.0k ★
Forks 6.3k
Category Other
Topics
Multi-AgentMicrosoftPythonCode ExecutionResearch
Ready to try AutoGen?
Free and open source. Install in one command, works with Claude Code, Cursor, Windsurf, and more.
$pip install pyautogen