LangGraph is LangChain's framework for building stateful, multi-actor AI applications modeled as directed graphs. While LangChain provides building blocks for LLM apps, LangGraph adds orchestration for complex agent workflows with cycles, branches, and persistence.
Workflows are modeled as graphs where nodes represent computation and edges define flow. Graphs support cycles for iteration and conditional branching based on intermediate results — capabilities simple chains lack.
Built-in persistence maintains state across interactions. Human-in-the-loop patterns add approval steps or corrections at any point. Streaming provides real-time visibility into execution.
Supports single-agent and multi-agent architectures with shared or independent state. LangSmith and LangGraph deployment surfaces provide production deployment, monitoring, and tracing around the open-source Python and JavaScript/TypeScript framework.
