Agent loop
The repeated cycle of a model observing state, deciding an action, calling a tool, and observing the result before deciding the next action, continuing until a task is done or a stopping condition is reached. Tool calling is what the model does inside one turn of the loop, not the loop itself.
Why exams ask this
Tested as "why did the agent stop after one step" or "why did it loop forever." The exam wants the observe-plan-act-observe cycle named directly, plus a termination condition, since a loop with no stopping rule is a failure mode, not a feature.
Relevant to
Related concepts
Resources
No resources linked to this concept yet.