Kamus AI
Agent loop
An agent's repeating cycle: the model picks a step, the app runs a tool, the result goes back to the model.
An agent loop runs until a stop condition holds: the task is done, the turn limit is reached, the budget runs out, or the agent needs to ask a person. Without clear stop conditions a loop can run long, burn money or repeat the same mistake. Every turn adds to the context, so the cost of one task grows with its number of turns.
Also called: Agent loop, agentic loop, turn
