★ OverviewBeginner10 min

What is LangGraph?

Nodes, edges, and the state machine model. Why LangGraph over plain chains.

Quick Reference

→Graphs beat chains — branches, cycles, and parallel execution are first-class
→StateGraph is the main entry point: define state type, add nodes, add edges, compile()
→compile() validates the graph and returns a CompiledGraph you can invoke or stream
→START and END are special pseudo-nodes marking entry and exit points
→Super-steps: each node execution is one super-step; the graph loops until END or recursion_limit

Why Graphs Beat Chains

Chains are linear — step A → step B → step C. Agents need more: branches (choose a path), cycles (loop back to retry), and parallelism (run steps concurrently). LangGraph models this as a state machine.

Chains are linear. Graphs branch, loop, and parallelize.

StateGraph: The Entry Point

StateGraph

StateGraph is the main class. The pattern: 1) define state type, 2) create graph, 3) add nodes, 4) add edges, 5) compile.

START, END, and the Execution Loop

▸START is a pseudo-node that marks the entry point of the graph — execution begins here.
▸END is a pseudo-node that signals termination — reaching END stops the graph and returns the final state.
▸Each node execution is one super-step. The graph processes one super-step at a time, updating state after each.
▸recursion_limit (default 25) caps the total number of super-steps to prevent infinite loops.
▸The graph loops through super-steps until it reaches END or hits the recursion_limit, whichever comes first.

Sign in to read this article

This is a premium article. Sign in with your Google account to continue.