AutoGen / AG2

Overview

AutoGen is an open-source programming framework for building AI agents and facilitating cooperation among multiple agents to solve tasks. It provides a high-level abstraction for creating conversable agents that can be configured to communicate with each other, use tools, and involve human feedback. Originally developed at Microsoft Research, AutoGen has evolved into AG2 under community-driven development.

Key Features

• Multi-Agent Conversations -- Agents communicate through structured conversations to solve complex tasks collaboratively.
• Human-in-the-Loop -- Seamless integration of human feedback and oversight within agent workflows.
• Tool Use -- Agents can execute code, call APIs, and use external tools to accomplish tasks.
• Customizable Agents -- Highly configurable agent behaviors, conversation patterns, and orchestration strategies.
• State-Driven Workflows -- StateFlow enables defining complex agent workflows as finite state machines for reliable execution.

Impact

AutoGen has gained significant adoption with over 50,000 GitHub stars and is used across industry and academia for applications ranging from software engineering and data analysis to scientific research and customer support. The framework introduced the concept of conversable agents, enabling a new paradigm for building LLM applications through multi-agent collaboration.

Key Publications

• AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation

  COLM 2024

• StateFlow: Enhancing LLM Task-Solving through State-Driven Workflows

  NeurIPS 2024 (Spotlight)

• AutoGen Studio: A No-Code Developer Tool for Building and Debugging Multi-Agent Systems

  EMNLP 2024 Demo