AutoGen
Build LLM-based multi-agent systems with patterns like GroupChat. Microsoft's original AutoGen repo is now in maintenance mode and community-managed; Microsoft names its own Agent Framework as the enterprise-ready successor for new projects, with AG2 available separately as an independently-maintained community fork.
The verdict on AutoGen: Developers building multi-agent systems that need iterative back-and-forth reasoning between agents AutoGen's real selling point, beyond the multi-agent patterns, is the human proxy: the ability to build supervised workflows where a human stays in the loop without writing custom interrupt logic from scratch. Pricing: Free and open source (in maintenance mode; Microsoft recommends new projects use Microsoft Agent Framework).
Best For
Developers building multi-agent systems that need iterative back-and-forth reasoning between agents
Standout Feature
The human proxy pattern makes it straightforward to build supervised, not fully autonomous, multi-agent workflows
TL;DR
Strong for complex research and coding tasks, now in maintenance mode, check whether the AG2 fork better fits new projects.
Alternatives
Fit Score: use-case reference
How this is scoredAutoGen scored 1 to 5 on five dimensions for the named use cases below. Part of our catalog-wide use-case reference, not a product-quality review rating.
Maintaining human-supervised research agent workflows
3.2/5- Capability
- 4/5
- Reliability
- 2/5
- Value
- 3/5
- Ease
- 3/5
- Compatibility
- 4/5
Keep iterative research agents working with a human review point in an existing AutoGen project. The human proxy pattern supports supervision, but the original repository is community-managed and receives no new features. Free code can preserve an existing investment, while model calls and maintenance still cost time and money; this is not a new-project recommendation.
Iterating code with agents in existing AutoGen projects
3.0/5- Capability
- 4/5
- Reliability
- 2/5
- Value
- 3/5
- Ease
- 2/5
- Compatibility
- 4/5
Support a coding loop in which agents write, test and revise code, with tool use and a listed execution sandbox. The code-first workflow needs engineering oversight, and a sandbox feature is not evidence that generated code is safe or correct. Maintenance-only status and ongoing model costs limit value for expansion, so evaluate the official successor before starting a new project.
Editorial judgment grounded in this tool's own published pricing, features, and our existing pros and cons, not a certification or a measured benchmark. Product quality is assessed separately under our review methodology.
Overview
Not yet reviewedProduct quality is assessed from original evidence. How reviews work.
AutoGen is Microsoft Research's open-source framework for building multi-agent AI applications where agents can converse with each other, use tools, execute code, and collaborate to solve problems that require multiple rounds of reasoning. The ConversableAgent pattern lets you define human proxies, coding agents, and assistant agents that hand tasks back and forth until a problem is solved, or escalate to a human when they're stuck. Used by researchers and developers building complex AI pipelines, automated research workflows, and coding agents that write, test, and iterate on code autonomously.
The source of the legacy catalog rating and review count has not been substantiated. Neither is presented here as a verified review rating or included in review structured data.
Workflow Fit Scores are separate, use-case-specific editorial judgments, not user-review averages. Read the Fit Score methodology.
- • Best for complex research and coding tasks that need iterative agent collaboration
- • Human proxy pattern makes it easy to build supervised autonomy workflows
- • Layered Core/AgentChat/Extensions architecture works for both quick prototyping and lower-level custom control
- • Higher complexity than simpler agent frameworks for basic tasks
- • Python-only with a steeper learning curve than visual tools
Key Features
- ConversableAgent pattern
- Human-in-the-loop support
- Code execution sandbox
- Group chat between agents
- Tool use and function calling
- Flexible model backend
Trust & Data
Verified 2026-09
- Training on your data
- Not publishedAutoGen is a self-hosted open-source Python framework with no SaaS layer, so no customer data is transmitted to Microsoft by design; no explicit data-governance statement exists on the project site or GitHub README regardless.
- Compliance
- Not published
- Retention
- Not published
- Export
- Not published
Sources: github.com, microsoft.github.io
As published by the vendor. Verify independently before purchase decisions.
Our Take
That makes it genuinely useful for research and coding tasks that need iterative back-and-forth between agents. The maintenance-mode status is the honest caveat: Microsoft's own README says the project will not receive new features and is now community-managed, and it names Microsoft Agent Framework as the enterprise-ready successor with stable APIs and long-term support. For a new project, check Microsoft Agent Framework first; AG2 exists as a separate, independently-maintained community fork worth comparing, not Microsoft's official path.