
LangGraph
MIT-licensed Python library for stateful agent workflows
- #2 of 12 in best AI agent frameworks and tooling
- Free and open source
- Free to use
- Open source
- Self-hosted
Ranked 2 of 12 in best AI agent frameworks and tooling
MIT-licensed graph runtime for agents, with durable state and human pauses, from the LangChain team. It sits under LangChain here as the lower-level layer, and LangSmith hosting is a separate paid service.
Best for Stateful agents that need explicit control of each step
About LangGraph
LangGraph models an agent as a graph. Nodes run model calls or tools, edges decide the next step, and state is kept between steps. The project describes durable execution, memory, streaming and human-in-the-loop pauses, and it is installed as a Python package with pip. Its repository is MIT licensed.
The library sits under LangChain's higher-level agent tools and suits teams that need explicit control over each step. The LangGraph FAQ says it is free to use. The paid part of the ecosystem is LangSmith, which covers observability, evaluation and deployment. Deep Agents, a higher-level package built on LangGraph, is an option for teams that want less setup.
Key features
- Durable execution
- Memory and streaming
- Human-in-the-loop
- Python package
- Deep Agents package on top
Strengths
- MIT-licensed library with durable execution
- Explicit state graphs for controlling each step
- Human-in-the-loop pauses built in
Limits
- Lower-level than most frameworks, so simple agents need more code
- Tracing and deployment need the paid LangSmith service
Where LangGraph ranks
How LangGraph compares
| Product | Score | Price | Platforms | Language | Agent approach | Paid hosted service |
|---|---|---|---|---|---|---|
| #1 | 92.0 | Free and open source | Self-hosted | Python and TypeScript | Graph workflows with LangGraph | LangSmith paid plans |
| #2 | 90.5 | Free and open source | Self-hosted | Python | Graph of steps with state | LangSmith paid plans |
| #3 | 88.0 | Free and open source | Self-hosted | Not listed | Document-centric agents | LlamaParse, separate |
| #4 | 86.5 | Free and open source | WindowsmacOSLinux | Python | Typed agents and structured outputs | Logfire, free tier |
| #5 | 85.5 | Free and open source | Command lineWindowsmacOSLinux | Python, TypeScript, Go and Java | Code-first agents with a web interface | Agent Runtime on Google Cloud |
The full comparison table for best AI agent frameworks and tooling has all 12 products.
Alternatives to LangGraph
All 11 alternatives
LangChainMIT-licensed framework for LLM apps and agents, with LangGraph
LlamaIndexMIT-licensed framework for document-based and agentic AI apps
Pydantic AIAlso runs on Windows, macOS, Linux.
Agent Development Kit (ADK)Also runs on Windows, macOS, Linux.
HaystackAlso runs on Windows, macOS, Linux.
CrewAIAlso runs on Windows, macOS, Linux.
Questions about LangGraph
- Is LangGraph free?
- LangGraph can be used without paying: MIT-licensed library is free; LangSmith hosting is a separate paid service. Check the vendor’s pricing page for current limits.
- Is LangGraph open source?
- Yes. LangGraph publishes its source code under the MIT license.
- Which platforms does LangGraph support?
- LangGraph is available on: Self-hosted.
- What is LangGraph best for?
- Stateful agents that need explicit control of each step. In our ranking of AI agent frameworks and tooling it is number 2 of 12 with a score of 90.5.
- What are the best alternatives to LangGraph?
- The highest-ranked alternatives on listme.name are LangChain, LlamaIndex, Pydantic AI, Agent Development Kit (ADK).
Show this ranking on your site
If you make or maintain LangGraph, this badge links back to this page and shows its current rank. It updates when the ranking does.
<a href="https://listme.name/software/langgraph"><img src="https://listme.name/badge/langgraph.svg" height="28" alt="Ranked #2 of 12 AI agent frameworks and tooling on listme.name"></a>