LangGraph is LangChain's low-level orchestration library for building stateful, multi-actor LLM applications as directed graphs. Nodes are functions (or agents), edges are transitions, and a persistent state object is threaded through every step.
When to reach for it
- You need cycles, retries, or human-in-the-loop checkpoints — not just a linear chain.
- You want durable execution: pause a run, resume it hours later, or rewind to a prior state.
- You're building production agents that need observability and time-travel debugging.
Key features
- Graph API — nodes, conditional edges, sub-graphs.
- Persistence — SQLite, Postgres, or Redis checkpointers.
- Human-in-the-loop — interrupt a graph, ask a human, resume.
- Streaming — token-, node-, and event-level streams.
Learn more
- Docs: <https://langchain-ai.github.io/langgraph/>
- GitHub: <https://github.com/langchain-ai/langgraph>