Definition
What is an agent loop?
An agent loop is the structure that lets an AI agent work unattended. On each cycle it reads the current state, finds the next piece of work, does it, has the result verified, writes down what happened, and decides the next move. The unit of work stops being the prompt and becomes the loop.
A loop differs from a cron job in one word: judgment. Anthropic’s guidance on building effective agents draws the same line between a fixed workflow and an agent that directs its own process. Cron runs a fixed script; a loop runs a model that reads state and chooses its next action. The load-bearing parts are a verifiable stopping condition (something a script can check, not “make it better”), a state file that persists between runs, and a budget ceiling so the loop cannot run up an unbounded bill.
Stub · This definition is being expanded with verified examples from shipped systems. Updated .
VibeMarketers publishes one verified marketing agent system every week, with the installable files.