Agent Loop

The core execution cycle of an AI agent: observe (receive input or trigger), think (reason about what to do using an LLM), act (call tools or take actions), and evaluate (check results and decide whether to continue or stop). Most agent frameworks implement this loop with configurable stopping conditions, retry logic, and maximum iteration limits to prevent runaway execution.