Autonomous AI agents execute iterative loops, alternating between model reasoning thoughts and external tool invocations to fulfill user tasks. Clicking Run Agent Trace dispatches each node across the timeline, validating JSON schemas and logging execution latency at every step. Observations feed back into the prompt context stack, accumulating token consumption until the agent synthesizes the final verified response.