Microsoft AutoGen Documentation
by Microsoft
Microsoft's event-driven multi-agent framework — now in maintenance mode, and worth reading as architecture.
Overview
AutoGen is Microsoft's open-source multi-agent framework — MIT-licensed code, roughly 60k GitHub stars — and its documentation mirrors the package split. autogen-core is the low-level event-driven runtime for message passing between agents, local or distributed, with cross-language support for Python and .NET. autogen-agentchat is the opinionated layer on top for rapid prototyping, providing two-agent chats, group chats and the other common team patterns. autogen-ext holds first- and third-party integrations, including LLM clients such as OpenAI and AzureOpenAI and capabilities such as code execution. Alongside the libraries, the docs cover AutoGen Studio, a no-code GUI for assembling and running agent teams, and AutoGen Bench for benchmarking. The AgentChat quickstart is about six lines: instantiate an AssistantAgent with an OpenAIChatCompletionClient, then await agent.run(task=...). Legacy 0.2 documentation is still linked separately, because the rewrite changed the API substantially and older tutorials will not run. The critical context the docs site does not lead with, but the repository README states plainly: AutoGen is in maintenance mode, receiving bug fixes, security patches and documentation improvements only, and Microsoft directs new projects to the Microsoft Agent Framework — the merger of AutoGen and Semantic Kernel that reached 1.0 general availability in April 2026. Treat this documentation as the reference for existing AutoGen systems and as a source of architectural ideas, not as the starting point for a new build.
At a Glance
- Topic
- Frameworks
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Self-paced reference across four surfaces — Studio, AgentChat, Core and Extensions
- Provider
- Microsoft
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Assemble agent teams using AgentChat's two-agent and group-chat patterns
- ✓Write event-driven agents directly on the autogen-core message-passing runtime
- ✓Attach tools, model clients and code executors through autogen-ext
- ✓Prototype an agent team in AutoGen Studio without writing code
- ✓Insert human-in-the-loop approval steps into an agent conversation
- ✓Run agents on a distributed runtime, including Python and .NET interop
- ✓Plan a migration path from AutoGen to the Microsoft Agent Framework
Highlights
- •Layered by design: use AgentChat for speed, drop into Core for control
- •Cross-language — the Core runtime supports both Python and .NET
- •AutoGen Studio gives non-programmers a working prototype surface
- •MIT-licensed code, ~60k GitHub stars, with 0.2 docs still archived
- •Read with the caveat: maintenance mode since 2026, superseded by Microsoft Agent Framework 1.0
Who It’s For
Best For
- ✓Teams maintaining or extending an existing AutoGen deployment
- ✓Engineers studying event-driven multi-agent architecture patterns
- ✓.NET shops that need agents alongside a Python runtime
- ✓Anyone planning an AutoGen to Microsoft Agent Framework migration
Prerequisites
- •Intermediate Python including asyncio — the whole API is async
- •An LLM provider key (OpenAI or Azure OpenAI) to run the quickstart
- •Awareness that new projects should evaluate Microsoft Agent Framework first
FAQ
What is Microsoft AutoGen Documentation?
The official documentation for Microsoft's AutoGen, split across four surfaces: AutoGen Studio (no-code prototyping UI), AgentChat (the high-level Python API for agent teams), Core (the event-driven message-passing runtime, with .NET support), and Extensions (model clients and capabilities such as code execution). Read it knowing AutoGen is now in maintenance mode.
Is Microsoft AutoGen Documentation free?
Microsoft AutoGen Documentation is free to access.
What level is Microsoft AutoGen Documentation for?
Microsoft AutoGen Documentation is aimed at a intermediate audience. Recommended background: Intermediate Python including asyncio — the whole API is async, An LLM provider key (OpenAI or Azure OpenAI) to run the quickstart, Awareness that new projects should evaluate Microsoft Agent Framework first.
How long does Microsoft AutoGen Documentation take?
Expect roughly Self-paced reference across four surfaces — Studio, AgentChat, Core and Extensions. Most learners work through it at their own pace.
What will I learn from Microsoft AutoGen Documentation?
You'll learn: Assemble agent teams using AgentChat's two-agent and group-chat patterns; Write event-driven agents directly on the autogen-core message-passing runtime; Attach tools, model clients and code executors through autogen-ext; Prototype an agent team in AutoGen Studio without writing code; Insert human-in-the-loop approval steps into an agent conversation; Run agents on a distributed runtime, including Python and .NET interop; Plan a migration path from AutoGen to the Microsoft Agent Framework.
Topics
Sources
This page was written from 2 sources, 1 on domains other than microsoft.github.io.