Agent graph (state graph)
Modeling an agent as explicit nodes, edges and shared state, with checkpoints, so each step can be paused, resumed, inspected and tested.
Instead of one open loop where the model decides everything, each node is a step (classify, retrieve, draft, ask a human) and edges say what may come next. LangGraph popularized the style, and other agent frameworks offer similar graphs. The state is saved after each node, which gives durable execution, approval steps and replay of a failed run from the last good point. It sits at the opposite end from a Ralph loop: more control and auditability, more design work up front.
Related terms
- Durable execution: Running a long process so its progress is saved after every step, letting it survive crashes and pause for hours or days, then resume.
- Agent framework (LangChain, LangGraph and others): A library for building AI agents and LLM workflows: model calls, tools, state, retries and tracing, so you do not write the loop yourself.
- Human in the loop: A person approves, edits or rejects the agent’s action before it takes effect.
- Ralph loop (Ralph Wiggum technique): Running the same prompt against a coding agent again and again, each time with a fresh context, while progress lives in files and tests decide when it is done.
