Skip to content

Agent Framework

langchain-ai/langgraph

langchain-ai/langgraph· Python

Graph-based orchestrator for stateful, durable LLM agents with checkpointing and human-in-the-loop.

GitHub stats

Stars
30,807
Forks
5,260
Watchers
154
Open issues
508

meta

License
MIT
Primary language
Python
Last commit
2026-04-29
Stats fetched at
2026-04-29

LangGraph models agent workflows as a graph of nodes and edges, with built-in state persistence, checkpointing, time-travel debugging, and human-in-the-loop interrupts. Built by the LangChain team for production multi-agent systems where you need explicit control over branching, retries, and long-running state. Install via `pip install langgraph`; pairs with LangSmith for tracing and LangGraph Platform for hosted deployment. Backs Anthropic's deep agents pattern and is widely adopted as the 2026 enterprise default.

Editor's verdict

Pick LangGraph when you need durable state, explicit control flow, and human approval gates — it beats CrewAI and AutoGen on production-readiness and debuggability. The trade-off is verbosity: defining nodes, edges, and TypedDict state feels heavy compared to a single-prompt ReAct loop, and you'll inherit the LangChain abstraction surface whether you want it or not. Skip it for one-shot chat or simple tool-calling — use the OpenAI Agents SDK or raw function calling instead.

Last updated: 2026-04-29

We use cookies

Anonymous analytics help us improve the site. You can opt out anytime. Learn more