AutoGen

framework
https://github.com/microsoft/autogen

AutoGen is Microsoft's open-source framework specifically designed for multi-agent conversations. Unlike general-purpose LLM frameworks, AutoGen was built from the ground up around the concept of agents talking to each other.

The framework provides primitives for agent-to-agent messaging, group chat with dynamic speaker selection, human-in-the-loop patterns, and code execution. AutoGen agents can be backed by different LLMs, allowing you to use the right model for each agent role.

AutoGen is particularly strong for research-oriented multi-agent systems where agent dialogue and debate are core to the workflow.

Key Features

  • Native multi-agent conversation support
  • Group chat with speaker selection
  • Human-in-the-loop integration
  • Code execution environment
  • Flexible agent configuration

Pros

  • +Purpose-built for multi-agent systems
  • +Microsoft research backing
  • +Clean conversation primitives
  • +Good for debate and verification patterns

Cons

  • Primarily Python
  • Smaller ecosystem than LangChain
  • Less mature tooling ecosystem