LangGraph, introduced by the LangChain team in 2024, is a library that models Agent workflows as a directed graph. Nodes represent LLM calls, tool calls, or human-approval steps; edges encode conditional transitions, and a shared state object travels through the run. Unlike "pure" LangChain, it embraces loops, branching, and human-in-the-loop pauses, which makes it a natural fit for Orchestrator designs. It has gained traction for production agents thanks to its persistence, durability, and observability features.
External Links