LangGraph is a low-level orchestration framework from LangChain Inc for developers building long-running, stateful AI agents, used by companies including Klarna, Replit, and Elastic. It can be used standalone or alongside other LangChain products for a fuller agent-building toolkit.
A developer installs LangGraph with pip and builds an agent as a workflow with persistent state, so execution can pause and resume exactly where it left off after a failure. Human-in-the-loop support lets a person inspect and modify agent state at any point during execution, and LangSmith provides visualization tools that trace execution paths and capture state transitions for debugging.
LangGraph provides durable execution for agents that run for extended periods, comprehensive memory with both short-term working memory and long-term persistent memory across sessions, and production-ready deployment infrastructure for stateful workflows. It pairs with Deep Agents for higher-level planning and subagents, and with LangSmith Deployment for scaling and sharing agents across teams.
LangGraph is open source under the MIT license, installed with pip install -U langgraph, and also available as LangGraph.js for JavaScript/TypeScript. It has more than 44.4k GitHub stars and 7.6k forks, and its design is inspired by Pregel, Apache Beam, and NetworkX.