AutoGen

What it does

AutoGen is a programming framework for building AI agents and multi-agent applications that act autonomously or alongside humans. Developers write code in Python (or .NET) that defines agents, tools, memory and model clients; agents exchange asynchronous messages to work through a task, and the output is whatever the application produces, such as agent conversations, tool calls and results.

It has layered APIs: a Core API for event-driven, distributed agents, a higher-level AgentChat API for patterns like two-agent and group chats, and an Extensions API for model clients, code execution and MCP tools. Built-in tracing supports OpenTelemetry. AutoGen Studio offers a no-code interface for prototyping workflows, and AutoGen Bench provides a benchmarking suite.

It suits developers and researchers studying or maintaining existing AutoGen code. The project is archived: Microsoft provides no bug fixes, security fixes, new features or support, and advises against production use. New projects are directed to Microsoft Agent Framework, with a migration guide for existing users. The code is free under the MIT license, but needs Python 3.10 or later, and examples call model providers such as OpenAI that require their own account and API key.

What it can do

• Build single and multi-agent applications with asynchronous, event-driven messaging
• Prototype common team patterns such as two-agent chat and group chat via AgentChat
• Connect tools and MCP servers to agents through extensions
• Trace and debug agent interactions with OpenTelemetry support
• Prototype workflows without code in AutoGen Studio
• Benchmark agent performance with AutoGen Bench

Works with

• OpenAI
• Azure OpenAI
• MCP servers (e.g. Playwright MCP)
• OpenTelemetry

Example workflows

Two-agent task prototype

Input: a task prompt and an OpenAI API key. Steps: install autogen-agentchat and autogen-ext[openai], create an AssistantAgent with a model client, run it on the task. Output: the agent's response printed from the run.

Web-browsing assistant via MCP

Input: a research question and the Playwright MCP server installed with npm. Steps: configure McpWorkbench with server parameters, attach it to an AssistantAgent, stream the run. Output: an answer built from tool-driven browsing.

Before you choose

• Project is archived with no security fixes; assess risks before handling sensitive data
• AutoGen Studio is a prototyping tool, not a production-ready app
• Only connect to trusted MCP servers, as they may execute local commands

Listed under

• Coding
• Agent framework
• AI agents
• Developers
• Developer tools
• Dotnet
• Framework

Get started

pip install autogen-agentchat; model provider key needed.

1. Install Python 3.10 or later
2. Run pip install -U "autogen-agentchat" "autogen-ext[openai]"
3. Export your model provider API key, then run a quickstart sample

Code is MIT-licensed. Running agents requires your own compute and model provider (e.g. OpenAI) account and usage charges, which are not priced here.

Source: github.com/microsoft/autogen, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• AutoGen - Microsoft Research ↗www.microsoft.com
• GitHub - microsoft/autogen: A programming framework for agentic AI · GitHub ↗github.com

Company website

microsoft.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.