LangGraph
AI Engineer path · lesson 79 of 124Walk a stateful agent graph, see how channel reducers merge state, route with conditional edges and cycles (with a recursion limit), and get durable execution: checkpoints, resume, and human-in-the-loop.
Key insight
The router read the state and branched — first to tools, then (after the result) to END. The tools → agent cycle is what makes it an agent rather than a one-shot chain, and explicit state is what makes it debuggable and resumable.
The parts of a LangGraph app
Sort each concept by the role it plays in the graph.
Tap a piece, then tap where it goes — or drag it. Tap the tray to take one back.
Does work
Drop items hereDecides what runs next
Drop items hereCarries data between steps
Drop items hereTest your knowledge
What does a conditional edge in LangGraph do?
The short version
LangGraph is a board game: the state is the board, nodes are squares that change it, and the dice (router) read the board to decide where you go next — sometimes looping back, sometimes landing on FINISH.
In more depth
LangGraph
LangGraph models an app as a state machine: nodes mutate a shared state object and edges decide what runs next. Unlike a chain, edges can be conditional (branch) and cyclic (loop). The classic agent is agent → router → (tools → agent)* → END: the router inspects state (did the model request a tool?) and routes accordingly. Explicit state gives persistence, checkpoints, human-in-the-loop, and replay.
Interview Questions
References & Further Reading
Love this visualization?
Sign up to track your progress, earn achievements, use the AI assistant, and access your personal learning dashboard.
100% free • No credit card required