langchain vs autogen
Two of the top projects in Agent frameworks and SDKs, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
langchain
Python framework for building LLM agents and apps from interoperable components
autogen
Python and .NET framework for multi-agent apps, now in maintenance mode
Verdict
langchain ranks higher, 88 to 30. Of the 4 score parts we measure for both, langchain leads on 4, autogen on none.
- Adoptionlangchain leads by 60 points, 97 to 37: a wide gap.
- Freshnesslangchain leads by 42 points, 100 to 58: a wide gap.
- Maintenancelangchain leads by 82 points, 86 to 4: a wide gap.
- Agent-readylangchain leads by 30 points, 30 to 0: a wide gap.
Pick langchain if you are
Python developers building LLM agents and applications in code
- Common interface across model providers, embeddings and vector stores
- Large integrations library covering tools, retrievers and vector stores
Pick autogen if you are
Developers with existing AutoGen code or prototyping multi-agent apps in Python
- Layered design: Core, AgentChat and Extensions APIs let you pick the abstraction level
- Core runtime supports local and distributed execution, with Python and .NET packages
| What we compare | langchain | autogen |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 97, widely used | 37, known |
| Freshness | 100, active | 58, recent |
| Maintenance | 86, healthy | 4, weak |
| Easy to run | Not scored for this kind | Not scored for this kind |
| Agent-ready | 30, minimal | 0, none |
| Facts from GitHub and the README | ||
| Stars | 147.8k | 61.4k |
| License | MIT (permissive) | CC-BY-4.0 (read the license) |
| Last commit | Oct 2026 | Apr 2026 |
| Last release | Oct 2026 | Sep 2025 |
| Language | Python | Python |
| Docker | No | No |
| GPU | Not stated | Not needed |
| arm64 or Apple Silicon | Not stated | Not stated |
langchain
LangChain is a Python library that gives a standard interface to chat models, embeddings, vector stores and tools, so providers can be swapped with little code change. Models load through init_chat_model with a provider:model string. For lower-level agent orchestration it points to LangGraph, and for tracing and evals to LangSmith.
Who it is for: Python developers building LLM agents and applications in code
Strengths
- Common interface across model providers, embeddings and vector stores
- Large integrations library covering tools, retrievers and vector stores
- Companions: LangGraph for orchestration, Deep Agents for planning and subagents
- MIT license; JS/TS counterpart available as LangChain.js
Weaknesses
- Observability, evals and managed deployment sit in LangSmith, a separate product
- Controllable stateful agent workflows need LangGraph, not core LangChain
- README gives no hardware requirements or list of supported providers
- Abstraction layer is another API to learn on top of provider SDKs
- Needs Model provider API or other model backend
- Models: OpenAI (README example), Other providers via integrations
autogen
AutoGen is a layered framework for building multi-agent applications: a Core API for event-driven message passing, an AgentChat API for two-agent and group-chat patterns, and an Extensions API for model clients, MCP tools and code execution. It also ships AutoGen Studio, a no-code prototyping UI on port 8080, and AgentBench. The project is in maintenance mode and community managed; Microsoft points new users to Microsoft Agent Framework.
Who it is for: Developers with existing AutoGen code or prototyping multi-agent apps in Python
Strengths
- Layered design: Core, AgentChat and Extensions APIs let you pick the abstraction level
- Core runtime supports local and distributed execution, with Python and .NET packages
- MCP tool support through McpWorkbench, e.g. the Playwright MCP server
- Includes AutoGen Studio for no-code prototyping and a benchmarking suite
Weaknesses
- Maintenance mode: no new features, only bug fixes, security patches and docs
- Maintainers recommend Microsoft Agent Framework for new projects
- AutoGen Studio is explicitly not production-ready and has no built-in authentication
- Support is community-only and response times may be slow
- no GPU
- Needs Python 3.10+, OpenAI API key (for the quickstart samples), Node.js/npx (for the Playwright MCP example)
- Models: OpenAI, Azure OpenAI
- port 8080
More in Agent frameworks and SDKs
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.