AutoGen is Microsoft’s open-source framework for building AI agents and multi-agent applications. It provides abstractions for agents that can communicate, use tools, execute workflows, and collaborate on complex tasks, allowing developers to move from simple assistants to coordinated systems involving multiple specialized roles. The project is aimed at developers experimenting with agentic application patterns as well as teams building more structured AI workflows in Python.

AutoGen includes components for defining agent behavior, message passing, tools, model clients, human participation, and event-driven multi-agent systems. Its modular design supports experimentation with different orchestration strategies rather than imposing one fixed agent architecture. It is best suited to developers who want programmatic control over how multiple AI agents reason, exchange information, and execute tasks, especially in research, prototyping, and custom application development.