AutoGen

AutoGen (Microsoft Research) is a multi-agent framework where agents communicate via structured conversations to solve tasks. It excels at code generation, execution, and iterative refinement through human-in-the-loop or fully autonomous agent teams.

Key features

• Conversable agents — every agent can both send and receive messages
• Code execution — agents can write and run Python/shell code in a sandboxed Docker environment
• Human proxy — seamlessly involve a human reviewer at any step
• Group chat — run round-robin or custom speaker-selection multi-agent chats
• Model flexibility — supports OpenAI, Azure OpenAI, Anthropic, local models via LiteLLM
• AgentChat API (v0.4+) — composable, async-first redesign with teams and handoffs

Quick start

npx ai-supply add autogen-multi-agent-conversations

# Or install directly
pip install pyautogen

import autogen

assistant = autogen.AssistantAgent(
    name="assistant",
    llm_config={"model": "gpt-4o", "api_key": "..."}
)

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    code_execution_config={"use_docker": False}
)

user_proxy.initiate_chat(
    assistant,
    message="Write and run a Python script that prints the Fibonacci sequence up to 100."
)

Curated mirror of the open-source AutoGen project (MIT). Install upstream from the repository.