Microsoft AutoGen

AutoGen is a Python framework for creating multi-agent AI applications that can operate autonomously or collaborate with humans, featuring a layered design for flexibility from high-level APIs to low-level components. It is primarily for researchers and developers experimenting with multi-agent orchestration patterns. However, it is now in maintenance mode, with Microsoft recommending their enterprise-ready successor, Microsoft Agent Framework, for new projects which offers stable APIs and long-term support. The community found its pioneering approach to multi-agent orchestration inspiring, despite its transition to maintenance mode, as it allowed for experimental multi-agent orchestration patterns and influenced subsequent frameworks like the Microsoft Agent Framework. While AutoGen is in maintenance mode and no longer receives new features, its architecture remains valuable for understanding multi-agent systems. Its core API implements message passing, event-driven agents, and local and distributed runtimes. Its AgentChat API provides a simpler, opinionated interface for rapid prototyping, and its Extensions API supports first- and third-party extensions for continuous capability expansion. These developer tools include AutoGen Studio, a no-code GUI for building multi-agent applications, and AutoGen Bench, a benchmarking suite for evaluating agent performance. This framework enabled the development of applications like Magentic-One, a state-of-the-art multi-agent team that can handle tasks requiring web browsing, code execution, and file handling.

Framework for multi-agent AI applications.