AutoGen (Microsoft)

About AutoGen (Microsoft)

AutoGen was created by Microsoft Research and released as an open-source project in 2023. The framework enables developers to build applications where multiple LLM-powered agents interact through conversations to complete complex, multi-step tasks. Unlike single-agent frameworks, AutoGen is designed around the concept of agents that can exchange messages, delegate tasks to each other, and involve human users in the conversation loop. AutoGen's agent model includes AssistantAgent (powered by an LLM, typically GPT-4 or another model), UserProxyAgent (which can execute code, run commands, and relay human input), and GroupChatManager (which orchestrates multi-agent conversations). This architecture enables patterns like code-generation-and-execution loops, research-and-writing pipelines, and complex problem-solving workflows that involve verification steps. The framework is MIT-licensed and can be used with any LLM provider that supports the OpenAI API format—including local models via Ollama, Azure OpenAI, Anthropic Claude, and EU-sovereign inference providers. This makes AutoGen deployable in fully sovereign EU environments without any dependency on Microsoft cloud services. Microsoft has continued to invest in AutoGen, releasing AutoGen 0.4 with a more modular architecture and introducing AgentChat as a high-level API. The project has become one of the most widely cited frameworks in multi-agent AI research and enterprise AI development. For European enterprises, AutoGen's MIT licence and LLM-provider flexibility make it a strong choice for building internal AI agent systems on EU-sovereign infrastructure. The framework itself creates no data residency obligations—those are determined entirely by the infrastructure and LLM providers chosen for deployment.