Extended thinking
The agent works through problems visibly before responding. You see a “Thinking…” indicator in real time, and when it finishes it shows how long it thought. You can expand the panel to read the full reasoning or collapse it. This makes the agent’s conclusions easier to trust and verify. It is especially useful for audits and strategic questions, where you want to understand the reasoning, not just the conclusion.Task planning
For multi-step requests, the agent creates a numbered plan before executing. Each step shows a status (pending, running, done, or failed) that updates as it works, so you see exactly what it is doing and in what order. For example:- Pull campaign performance, done
- Analyze ad set spend efficiency, running
- Review creative performance, pending
- Identify optimization opportunities, pending
- Build final recommendations, pending