langchain-ai/langgraph

Supports durable execution with automatic state persistence across failures, human-in-the-loop interrupts for agent inspection/modification, and comprehensive memory management combining short-term working state with long-term persistence. Built as a low-level orchestration framework inspired by Pregel and Apache Beam, it integrates seamlessly with LangChain ecosystem tools including LangSmith for observability and LangSmith Deployments for production scaling of stateful workflows.

26,286 stars and 42,304,147 monthly downloads. Used by 67 other packages. Actively maintained with 145 commits in the last 30 days. Available on PyPI.