AutoGen is an open source framework from Microsoft Research for building applications with multiple conversational AI agents. It lets developers define agents that talk to each other and to humans, coordinating to solve tasks through structured conversations. Agents can be given roles, tools, and the ability to execute code, enabling patterns such as an assistant agent working alongside a user proxy. The framework handles the messaging and orchestration between agents, so developers focus on defining behavior rather than managing the flow of turns manually.
The framework is aimed at researchers, developers, and technical teams exploring multi agent systems and complex LLM workflows. It fits people building code generation assistants, automated problem solving pipelines, and experiments that require agents to critique, plan, and collaborate. Because it is code first and Python based, it appeals to those comfortable programming and integrating tools. Its research roots make it popular for prototyping novel agent interaction patterns, while still being usable for practical applications that benefit from dividing work across cooperating agents.
In practice, developers instantiate agents, assign roles and tools, and start a conversation that the framework coordinates automatically. Agents can generate and run code, call functions, and involve a human when needed. Developers adjust prompts and configurations to shape how agents reason and hand off tasks.
