LangGraph (multi-agent orchestration)

A library for building stateful, multi-agent applications by modeling workflows as cyclic graphs.

LangGraph provides the low-level orchestration needed to move beyond simple linear chains into complex, circular agentic workflows. By using a StateGraph architecture, you define nodes as specific functions (like a researcher or a coder) and edges as the logic that routes data between them. This framework excels at handling the messy realities of production: it manages persistent state across long-running sessions, supports human-in-the-loop interruptions for manual approvals, and enables error recovery through built-in checkpointing. Whether you are coordinating a supervisor agent to delegate tasks to specialized workers or building a self-correcting RAG pipeline, LangGraph gives you granular control over every state transition without the constraints of a black-box system.

https://langchain-ai.github.io/langgraph/