AutoGen is a framework from Microsoft Research for building agentic apps and multi-agent systems, with a developer SDK plus optional AutoGen Studio UI for working with and experimenting with agents.
AutoGen is a framework from Microsoft Research for building “agentic” applications, including systems where multiple agents coordinate to complete tasks.
In the v0.4 “stable” documentation, AutoGen is presented as a set of components rather than a single monolith: AgentChat (for conversational / agent interaction patterns), Core (for event-driven multi-agent systems), and Extensions (for adding capabilities), plus an optional AutoGen Studio experience.
AutoGen’s docs emphasize a messaging / runtime approach in Core (agents communicate via an agent runtime) and provide separate “user guides” for each component so you can choose the level of abstraction you want.
AutoGen is aimed at developers and teams building agentic applications (including multi-agent systems) who want an official Microsoft Research framework with both SDK components and optional Studio tooling.
AutoGen’s “stable” docs position it as a component-based framework for agentic apps, where you choose between higher-level chat patterns (AgentChat) and lower-level event-driven multi-agent systems (Core).
AgentChat focuses on building agent interaction and chat-oriented patterns, while Core documents an agent runtime/message model for building event-driven multi-agent systems.
Extensions add capabilities to the ecosystem, and AutoGen Studio provides a UI experience documented separately from the core SDK/framework pieces.
AutoGen’s official documentation and Microsoft Research project page present AutoGen as a framework/ecosystem; they do not present an official “pricing” page in the sources above. For licensing/cost, the docs link to the project’s GitHub repository as the canonical code source.
Open-source (MIT license) framework (no official pricing page referenced in the docs above)
AutoGen is presented as a Microsoft Research framework with distinct components (AgentChat/Core/Extensions/Studio). LangChain is a separate ecosystem (framework + optional platform products). If you want Microsoft’s component split and Studio option, AutoGen is the closer match; if you want the LangChain/LangGraph/LangSmith stack, that’s a different toolchain.
Learn more about LangChainAutoGen’s docs emphasize a component split (AgentChat/Core/Extensions/Studio) and an event-driven multi-agent “Core” runtime; CrewAI is commonly described as a Python multi-agent framework with “crews” and tasks (different abstraction). Choose based on whether you want AutoGen’s component architecture vs CrewAI’s “crew” model.
Learn more about CrewAILast updated: December 2025
AI Agent Framework