AutoGen from Microsoft Research enables multi-agent conversations. Agents have configurable behaviors. Human-in-the-loop supported naturally. Code execution within agent workflows.
Agent Types
AssistantAgent powered by LLMs. UserProxyAgent represents human input. Custom agents for specific behaviors. GroupChat manages multi-agent conversations.
- Configure assistant agents with system messages
- Use user proxy for human interaction
- Enable code execution for programming tasks
- Set up group chat for multi-agent discussion
- Customize termination conditions
Conversation Patterns
Two-agent conversations for simple tasks. Group chat for complex collaboration. Sequential chats for pipelines. Nested chats for hierarchical workflows.