LangGraph Guide
2026-08-01 · 6 min read
How graph designs map onto LangGraph in Python and TypeScript: state, nodes, conditional edges, and checkpoints.
Mapping the primitives
LangGraph models a workflow as a state machine. Your graph's shared state becomes the state schema, nodes become callables, and routing becomes conditional edges.
- Shared state → typed state schema
- Node → node function returning a state patch
- Edge → add_edge / add_conditional_edges
- Approval → interrupt before a node
Exports are scaffolds
Loop Engineering AI exports generate structured scaffolds and implementation guidance — not deployed applications or executed workflows. You still own credentials, tools, and deployment.