LangGraph

An open-source framework for building stateful, multi-step AI agents as graphs — for when a single prompt isn't enough.

What it is

A way to wire LLM calls, tools, and decisions into an explicit graph with memory, so an agent can loop, branch, and retry instead of answering in one shot.

When to use it

Multi-step agents: research-then-write, call-a-tool-then-decide, any workflow where the model needs several turns and has to keep state between them.

When NOT to

A plain chatbot or a single API call doesn’t need it, the graph is pure overhead until you actually have branching steps. Start simple; reach for LangGraph when the logic gets real.

Used in

The orchestration layer for agentic versions of Build an AI Chatbot.