Glossary
Agent loop
The cycle in which a model proposes an action, the harness runs an allowed tool, and the result comes back until a stop condition.
An agent loop is the control loop from goal to action. The model proposes a step. The harness runs an allowed tool, returns the result, and the cycle continues until the goal is met, a limit hits, or a person must approve.
Anthropic’s Form S-1/A, filed November 2, 2026 and read November 8, 2026, defines the agent loop as the cycle in which an agent receives a task, takes an action, observes the result, and decides the next step until the task is complete. The Claude Agent SDK and OpenAI’s Agents SDK, both read the same day, describe that cycle in a product runtime. A 2022 ancestor of the written form is ReAct.
The long explainer is How do AI agents work?. The workshop around the loop is What is an AI agent harness?. Choosing the next specialist is orchestration.