AutoGen, Archived by Microsoft: What the Framework Still Offers and Where It Falls Short
A programming framework for agentic AI
At a glance
- What is it?
- Microsoft's AutoGen is a Python framework for multi-agent AI applications, now archived and no longer maintained by Microsoft. This review covers its layered architecture, quickstart path, and why new projects should look at Microsoft Agent Framework instead.
- Who is it for?
- Adopt AutoGen if you have an existing v0.2 codebase and need a frozen framework for experimental multi-agent patterns and accept that Microsoft provides no fixes or support. Do not start new projects on it; the README explicitly directs new users to Microsoft Agent Framework.
- Can I use it commercially?
- Yes, with credit. CC-BY-4.0 allows commercial use as long as you credit the authors and indicate what you changed. It is written for creative content, so check how it applies to any code.
- Is it still maintained?
- Yes. The repository last received commits 3 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 13, 2026, and from our analysis. They are not legal advice.
Star history
- microsoft/autogen61,803 ★ · +1,310
| Repository | 2026-08-18 | 2026-11-13 |
|---|---|---|
| microsoft/autogen | 60.5k | 61.8k |
1,310 Stars in the last 87 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The Core Problem: Multi-Agent Orchestration Without a Production Guarantee
AutoGen solves the problem of coordinating multiple AI agents that can act autonomously or work with humans. It is a programming framework, not a hosted service, so you write Python code that defines agents, their tools, and how they interact. The intended users are developers who want to prototype multi-agent workflows, whether that is a simple two-agent chat or a group chat with specialized roles. The README is blunt about the project's status: AutoGen is archived and no longer actively maintained by Microsoft, which provides no bug fixes, security fixes, new features, or support. That is a decisive constraint. If you need enterprise support or long-term API stability, this is not the framework for you. But if you are exploring agent patterns and can tolerate a frozen feature set, it still provides a working foundation.
Layered Architecture: Core, AgentChat, and Extensions
The framework is divided into three layers with clear responsibilities. The Core API handles message passing, event-driven agents, and both local and distributed runtimes. It also supports cross-language development for .NET and Python. On top of that sits the AgentChat API, which is described as simpler but opinionated. It is the layer closest to what v0.2 users know, and it supports common patterns like two-agent chat and group chats. The Extensions API provides pluggable implementations for LLM clients, including OpenAI and AzureOpenAI, plus capabilities like code execution. This layering means you can work at a high level with AgentChat for rapid prototyping, or drop down to Core for finer control. The trade-off is complexity: you need to understand which layer you are using and how they interact, especially when you start mixing custom tools or distributed runtimes.
Getting Started: Installation and the Hello World Example
Installation requires Python 3.10 or later. The README gives two install commands. The first installs the AgentChat package and the OpenAI extension: pip install -U "autogen-agentchat" "autogen-ext[openai]". The second installs AutoGen Studio, the no-code GUI: pip install -U "autogenstudio". The quickstart shows a minimal agent: you create an OpenAIChatCompletionClient with a model name like "gpt-4.1", pass it to an AssistantAgent, and call agent.run with a task. The example requires an OPENAI_API_KEY environment variable. The code is async, so you wrap it in asyncio.run. That is the entire setup for a basic agent. There is no configuration file for this path; you set the model client in code. For more complex setups, you would need to consult the AgentChat documentation, which is linked but not reproduced in the README.
MCP Support and Multi-Agent Orchestration: Two Concrete Examples
The README shows two more advanced patterns. The first is an MCP server integration. You install the Playwright MCP server globally with npm install -g @playwright/mcp@latest, then use StdioServerParams to launch it via npx. The McpWorkbench class connects to that server, allowing an agent to control a browser. This is a practical way to give agents web browsing capabilities without writing browser automation code yourself. The warning in the README is important: only connect to trusted MCP servers because they can execute commands in your local environment or expose sensitive data. The second pattern is multi-agent orchestration using AgentTool. You wrap an AssistantAgent in an AgentTool and pass it to another agent, effectively letting one agent call another as a tool. The example defines a math_expert agent with a system message and description, then uses AgentTool to expose it. This is a simple form of orchestration, but it shows the core mechanism: agents as tools. For more advanced workflows, the README points to the AgentChat documentation.
AutoGen Studio: A No-Code GUI with Explicit Non-Production Warnings
AutoGen Studio is a separate package that runs as a local web app. The command is autogenstudio ui --port 8080 --appdir ./my-app, and it launches on http://localhost:8080. It lets you prototype multi-agent workflows without writing code. The README is unusually candid about its limitations: it is meant for rapid prototyping and to demonstrate an example of end user interfaces, but it is not a production-ready app. The security note says developers should implement authentication and other features themselves when building deployed applications. This is a real limitation. If you are evaluating AutoGen Studio as a potential product foundation, the answer is no. It is a prototyping tool, and the project itself tells you not to use it in production. That honesty is refreshing, but it also means you cannot skip the engineering work if you want to ship something real.
Maintenance Mode and the Migration Path to Microsoft Agent Framework
The most significant fact about AutoGen is its maintenance status. The README contains a caution box at the top: the project is archived and no longer actively maintained by Microsoft, with no bug fixes, security fixes, new features, or support, and the README advises against production use. New users are told to start with Microsoft Agent Framework (MAF), and existing users are encouraged to migrate using a dedicated guide. The README describes MAF as the enterprise-ready successor, with stable APIs and long-term support. It also mentions cross-runtime interoperability via A2A and MCP. This is not a subtle hint; the project is actively steering users away. For a technology editor, this changes the evaluation. You are not comparing AutoGen to other agent frameworks on features alone. You are comparing it to its own successor. The migration guide is the concrete next step for existing users, but the README does not detail what changes are required. You would need to read that guide to know the cost of moving from v0.2 or v0.7 to MAF.
Limitations and Wrong Use Cases: What the README Does Not Promise
The README does not hide the framework's boundaries. AutoGen Studio is explicitly not production-ready. The MCP warning highlights security risks. The archived status means no new features, bug fixes or security fixes, so any limitation you find today will likely stay. One clear wrong use case is a production deployment that needs long-term vendor support. Another is a project that requires new model integrations beyond what the Extensions API currently provides; since the project is frozen, you cannot expect new clients to be added. If you need a framework with an active roadmap, AutoGen is the wrong tool. If you are building a small internal prototype or learning multi-agent patterns, it can work, but you should budget time for a future migration to MAF.
Licensing and Upgrade Cost: CC-BY-4.0 and the v0.2 Migration Guide
The repository is licensed under CC-BY-4.0, which is unusual for a code framework. That license is typically used for creative works, not software. It allows sharing and adaptation with attribution, but it is not a standard software license like MIT or Apache 2.0. You should review what that means for your use case, especially if you plan to embed the code in a commercial product. The repository also has a LICENSE-CODE file next to LICENSE, so check both license texts directly. On upgrades, the README points to a migration guide for users upgrading from v0.2. The current stable version is 0.7.5, released on 2025-09-30. The release notes are not included in the README, so the exact changes between 0.7.4 and 0.7.5 are unknown. The archived status means no further upgrades or fixes are coming, so the upgrade cost is mainly the eventual migration, which is potentially high if you are on an old version like v0.2.
Editorial conclusion
Adopt AutoGen if you have an existing v0.2 codebase and need a frozen framework for experimental multi-agent patterns and accept that Microsoft provides no fixes or support. Do not start new projects on it; the README explicitly directs new users to Microsoft Agent Framework. Before adopting, verify that your model provider is supported by the Extensions API and check the migration guide if you are coming from v0.2. The archived status means no fixes or new features, so plan for a future migration to MAF if your project outgrows AutoGen's current capabilities.
Frequently asked questions
What is AutoGen used for?
AutoGen is a framework for creating multi-agent AI applications that can act autonomously or work alongside humans. Its layered API covers message passing, event-driven agents, a local and distributed runtime, and a simpler AgentChat layer for rapid prototyping.
Is AutoGen discontinued?
Effectively yes: the README says the project is archived and no longer actively maintained by Microsoft, with no bug fixes, security fixes, new features, or support, and new users are pointed to Microsoft Agent Framework as the enterprise-ready successor.
how to install autogen
AutoGen requires Python 3.10 or later. The stable install pulls the AgentChat package plus the OpenAI extension extra, while AutoGen Studio is a separate package installed on its own.
how to use autogen studio
Run autogenstudio ui --port 8080 --appdir ./my-app to serve the no-code GUI on http://localhost:8080. The project cautions that Studio is for rapid prototyping and is not a production-ready app, and that you should build your own application with authentication and security for deployment.
how to use mcp with autogen
The MCP example combines McpWorkbench with StdioServerParams pointed at the Playwright MCP server over npx. The project warns that you should connect only to trusted MCP servers, because they may execute commands in your local environment or expose sensitive information.
Official sources
Where this project is recommended
Where developers are discussing it
Posts on Hacker News, dev.to and Lobsters that link to this repository, found by our scan of those communities. 1 in the last 30 days; newest first.
- The Microsoft Lab That Worked on AutoGen Is Over. AutoGen Already Gets No Security Fixesdev.to · Nov 7, 2026 · 2 points · 0 comments
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/microsoft-autogen)Star history chart for your README
[](https://hysenlabs.com/projects/microsoft-autogen)Community discussions
Was this useful?