AutoGen

Microsoft's multi-agent conversation framework (now in maintenance mode).

Hand-reviewed by the Toolglade team·Updated Aug 2026

coding#agents#multi-agent#open-source#microsoft

Free plan API Self-hosted

Toolglade’s take

AutoGen is genuinely influential in multi-agent design, but honesty requires flagging that Microsoft moved it into maintenance mode in early 2026 and now steers new work toward the merged Microsoft Agent Framework. Existing AutoGen apps keep working and getting bug fixes, and it remains a fine way to learn multi-agent patterns, but teams starting fresh should seriously evaluate the successor framework before committing.

About AutoGen

AutoGen is Microsoft's open-source framework for building systems where multiple AI agents converse to solve tasks, with a 0.4-era asynchronous, layered architecture. It is free and self-hostable, but as of 2026 it is in maintenance mode, with Microsoft directing new development to the successor Microsoft Agent Framework that merges AutoGen and Semantic Kernel.

AutoGen pioneered the idea of framing agentic AI as a conversation between multiple specialized agents, for example a coder agent and a critic agent exchanging messages in a loop until a task is complete. The 0.4 rewrite introduced an asynchronous, event-driven, layered architecture (core, AgentChat, and extensions) intended to be more scalable and observable than the original 0.2 design. It is open source, works with multiple model providers, and has been popular in research and prototyping. The important 2026 context is that Microsoft has consolidated its agent efforts. AutoGen and Semantic Kernel have converged into the new Microsoft Agent Framework, which reached a 1.0 release for Python and .NET in April 2026 and is now the recommended production path for new Microsoft/Azure agent projects. AutoGen itself has moved into maintenance mode: it still receives bug fixes and community contributions, and existing 0.4 / 0.7.x deployments keep running, but net-new feature work is happening in the Agent Framework. As a result, AutoGen remains a usable and well-documented framework, especially for teams already invested in it or studying multi-agent patterns, but new projects should weigh whether to adopt the Microsoft Agent Framework instead to avoid building on a codebase that is no longer the strategic focus.

TL;DR

AutoGen is Microsoft's open-source framework for multi-agent AI systems built around agent conversations. It popularized patterns like coder-and-critic agent loops and shipped a scalable 0.4 rewrite. As of 2026 it is in maintenance mode, with new development directed to the successor Microsoft Agent Framework. It remains free and usable, especially for existing deployments and research. New projects should evaluate the Agent Framework before committing.

Company overview

AutoGen originated from Microsoft Research as a framework for agentic AI and became one of the most cited multi-agent toolkits. It is open source and community-supported alongside Microsoft's own teams.

In 2025 and 2026 Microsoft consolidated its agent tooling, merging AutoGen and Semantic Kernel into the Microsoft Agent Framework, which is now the supported production path.

Product features

AutoGen provides agent abstractions, group-chat and conversation orchestration, tool and code execution, and a layered 0.4 architecture (core runtime, AgentChat high-level API, and extensions). It supports multiple model providers and integrates with Azure.

Many of these ideas carry forward into the Microsoft Agent Framework, which adds enterprise features like session state management, middleware, telemetry, and graph-based workflows.

Target market

AI researchers and developers building or studying multi-agent systems, historically strong in prototyping and academic contexts, with Microsoft/Azure-aligned enterprises now pointed toward the Agent Framework.

Buyer personas

End users

AI engineers and researchers writing multi-agent Python (and .NET) code.

Buyers

Engineering leads at organizations evaluating agent frameworks, often within Microsoft/Azure environments.

Key influencers

Microsoft Research, open-source contributors, and AI developer community voices.

Ideal customer profile

Teams experimenting with or already running multi-agent systems who are comfortable with a code-first, open-source framework and aware of the migration path to the Microsoft Agent Framework.

Funding & performance

AutoGen is a Microsoft open-source project rather than a standalone company; it is funded and maintained by Microsoft. No separate startup funding applies.

Pros & cons

Pros

Pioneered accessible multi-agent conversation patterns
Free and open source
Backed by Microsoft Research with strong documentation
0.4 architecture is asynchronous and more observable
Works with many model providers
Large community and example library

Cons

In maintenance mode as of 2026, no new feature focus
Microsoft steers new projects to the Agent Framework
Multiple version lines (0.2 vs 0.4/0.7) cause confusion
Multi-agent loops can be hard to control and cost-predict
Less enterprise tooling than the successor framework
Migration uncertainty for long-term projects

Pricing plans

Open Source

Full framework, MIT-licensed
Self-host anywhere
Multi-agent orchestration
You pay only for LLM API and infrastructure

Key features

API

Self-hosted

Multi-language

Integrations

OpenAI, Azure OpenAI, Anthropic, local models, Microsoft Agent Framework, Python and .NET

Input types

text, code

Output types

text, code

Best For

multi-agent conversations, research prototyping, code-generation agents, agent collaboration patterns

Compare key features

View all alternatives →

Feature

AutoGen

Cline

vLLM

Pricing

Free

Free plan

Yes

Free trial

API

Yes

Self-hosted

Yes

Frequently asked questions

Is AutoGen still maintained?+

Yes, but in maintenance mode as of 2026. It receives bug fixes and community contributions, while new feature development has moved to the Microsoft Agent Framework.

What is replacing AutoGen?+

The Microsoft Agent Framework, which merges AutoGen and Semantic Kernel and reached a 1.0 release for Python and .NET in April 2026, is Microsoft's recommended path for new agent projects.

Is AutoGen free?+

Yes, AutoGen is open source and free. You only pay for the underlying LLM API usage and your own hosting.

Should I start a new project on AutoGen?+

For learning multi-agent patterns it is fine, but for new production projects you should evaluate the Microsoft Agent Framework first, since AutoGen is no longer the strategic focus.

What is the difference between AutoGen 0.2 and 0.4?+

The 0.4 rewrite introduced an asynchronous, event-driven, layered architecture (core, AgentChat, extensions) for better scalability and observability compared to the older 0.2 design.