LangGraph
A low-level orchestration framework for building long-running, stateful agents, with durable execution that resumes after failures, human-in-the-loop control, and short- and long-term memory.
LangGraph provides low-level infrastructure for any long-running, stateful agent or workflow. It supports durable execution so agents can persist through failures and resume from where they left off, human-in-the-loop hooks for inspecting and modifying agent state during a run, and both short-term working memory and long-term memory across sessions.
It integrates with LangSmith for tracing execution paths, capturing state transitions and evaluating agent trajectories, and with LangChain for reusable integrations and components. Deep Agents is a higher-level package built on top of LangGraph for agents that plan, use subagents and work with file systems. An equivalent JavaScript and TypeScript library, LangGraph.js, is also available.
It installs from PyPI, is written in Python, and is licensed under the MIT license.