AutoGen

AutoGen is Microsoft's multi-agent framework, centered on agents that communicate via chat. Define agents, they send messages to each other, problems get solved through conversation. Research-flavored but increasingly production-ready.

Core concepts

• ConversableAgent, agents that can send and receive messages
• UserProxyAgent, agent that can execute code and call tools on behalf of the conversation
• GroupChat, multiple agents in a shared chat with a designated manager

Strengths

• Elegant for problems that fit the "agents talking" metaphor
• Strong code execution support
• Built-in patterns for various multi-agent shapes
• Good for research and experimentation

Weaknesses

• Steeper learning curve than CrewAI
• Chat-message paradigm can be inefficient for simple workflows
• Less opinionated than some alternatives

When to pick AutoGen

• Research settings exploring multi-agent interaction
• Problems genuinely structured as agent-to-agent conversation
• Teams with Python + Microsoft tooling fit

AutoGen Studio

Microsoft offers AutoGen Studio, a UI for building and debugging agent configurations. Useful for non-programmers building agent workflows.