Concept
Agent Loop
The repeated think/act/observe cycle by which an LLM agent makes progress on a task.
Definition
The agent loop is the core control structure of an autonomous LLM system: it observes state, calls the LLM to plan, executes an action, observes the result, and repeats until done. Termination conditions and safety checks are critical for predictable behaviour.
Common use cases
- Autonomous agents
- Browser bots
- Coding agents