Skip to main content

COMPEL Glossary / state-graph

State graph

An agent execution modeled as a directed graph of states and transitions — each node is a reasoning or tool-call step, each edge is a conditional transition.

What this means in practice

State graphs make agent control flow explicit and auditable, enabling deterministic replay and targeted intervention.

Synonyms

agent state graph , LangGraph state graph

See also

  • Agent runtime — The execution substrate that hosts the agent loop, its tool calls, state, and recovery logic — e.g., LangGraph, CrewAI, AutoGen, or OpenAI Agents SDK.
  • Agent loop — The iterative pattern in which a model plans, calls tools, observes the results, and replans — continuing until a goal is met, an iteration budget is exhausted, or a human intervenes.
  • ReAct — An agentic prompt pattern interleaving reasoning steps (Thought) and action steps (Act — typically a tool call) over multiple turns, with observations from each action feeding the next reasoning step.
  • Replayability — The ability to deterministically reproduce an agent trace given captured inputs — a recorded session, the prompt, the tool responses, the random seed.