AutoGen

What is AutoGen?

AutoGen is a framework for building AI agents and applications. It is organized in three layers: AgentChat for conversational single- and multi-agent apps, Core for event-driven multi-agent systems, and Extensions for integrations with outside services and libraries.

Key capabilities of AutoGen

• AgentChat: a Python framework for single- and multi-agent conversational apps, requiring Python 3.10 or later.
• Core: an event-driven framework for scalable multi-agent systems and distributed agents.
• AutoGen Studio: a web-based UI for prototyping agents without writing code.
• MCP support: connects to MCP servers through McpWorkbench.
• Code execution: runs code in Docker-based environments.
• Distributed runtimes: supports gRPC-based distributed agent runtimes.

How AutoGen works

Developers install AutoGen with pip and compose agents in AgentChat, or drop down to Core for custom event-driven workflows. Extensions connect the agents to MCP servers, OpenAI Assistant agents and Docker code execution. AutoGen Studio, installed with pip install -U autogenstudio, offers a visual way to prototype.

Who uses AutoGen?

AutoGen is for developers and researchers building multi-agent systems, including deterministic and dynamic workflows and distributed agents. A .NET version is also linked from the documentation.

AutoGen pricing

AutoGen is an open-source framework that you install yourself, so there is no license fee listed. Your costs come from the model APIs and infrastructure you connect to it.

AutoGen alternatives

Toolhouse provides managed tools and infrastructure for agents rather than an orchestration framework. Smithery is a registry for MCP servers that agents can connect to. Nango handles integrations and authentication for AI agents. AutoGen differs by focusing on how multiple agents converse and coordinate.