What it does
LangGraph is described as a low-level orchestration framework for building, managing and deploying long-running, stateful agents. It is installed as a Python package with pip, and a separate JS/TS library, LangGraph.js, exists for JavaScript. It supplies supporting infrastructure rather than ready-made agent behaviour: execution that survives failures and resumes where it stopped, ways for a person to inspect and change agent state during a run, and both short-term working memory and long-term memory across sessions. It can be used on its own or alongside other LangChain products. LangSmith helps with tracing, evals and observability, LangSmith Deployment covers hosting, and Deep Agents is a higher-level package built on LangGraph for agents that plan and use subagents. LangGraph is built by LangChain Inc and can be used without LangChain.
Capabilities
- Durable execution that resumes after failures
- Human-in-the-loop inspection and editing of agent state
- Short-term and long-term memory
- Visibility into execution paths and state changes through LangSmith
- Deployment path through LangSmith Deployment
When to use it
- Building long-running agents that need persistent state
- Workflows where you want low-level control over how an agent runs
- Agents that need human oversight at points during a run
