microsoft/autogen

AutoGen is a Python framework from Microsoft for building applications where multiple AI agents work together to complete tasks. Rather than using a single AI model to handle everything, AutoGen lets you create a team of specialized agents that can communicate with each other, use tools, and coordinate to solve problems that are too complex for one agent alone. For example, you could have a coding agent that writes Python code, a testing agent that checks if the code works, and an orchestrating agent that manages the workflow between them. The framework is organized in layers. The core layer provides the fundamental building blocks for agents and messaging. Above that is AgentChat, a higher-level API that makes common patterns like conversation loops, multi-agent group chats, and tool use easy to implement. Agents can use MCP servers as tool providers, which is a standard protocol for connecting AI systems to external services like web browsers, databases, and APIs. AutoGen Studio is a no-code graphical interface for prototyping multi-agent workflows without writing code. The project is built in Python and supports integration with AI models from providers including OpenAI, Azure OpenAI, and others. The README prominently notes that AutoGen has been placed into maintenance mode and will not receive new features going forward. Existing users are encouraged to migrate to Microsoft's newer Agent Framework, which the documentation describes as the enterprise-ready successor. New users starting projects today are directed to that newer framework instead. You would encounter AutoGen when maintaining an existing project built on it, learning multi-agent concepts from its documentation, or studying the design patterns that influenced its successor.