How thinking is detected

Every money rule in Awaitful hangs on one phrase: your agent is genuinely thinking. This page explains what detection actually observes - and, just as important, what it never looks at.

What detection watches

Detection watches the agent's own thinking indicator - the spinner Claude Code itself shows while it works - and nothing else. It does not read what the agent is doing, what you asked it, or what it answers. The signal is binary and shallow on purpose: thinking started, thinking stopped. That an agent is thinking, and for how long, is the entire observation.

Which agents today

Claude Code is the only supported agent today, in VS Code, Cursor, Devin, VSCodium, and VS Code Insiders. The client is built with a seam for more agents, and the honest per-platform status lives on the landing page's platform table - nothing here will claim an agent before it ships.

What starts and stops the clock

Thinking starting is what makes the sponsored line appear; thinking stopping is what makes it vanish. In between, the visibility clock of What counts as an impression only accrues while the thinking is genuine: the agent stopping, a permission prompt waiting on you, or your window losing focus each pause the clock. A prompt waiting for your input is you working, not the agent - so it does not count, and the ad hides.

Why it is built this way

Detecting more would enable more - smarter targeting, richer analytics - and every one of those features would cost the only thing this product cannot spend: the certainty that nothing of yours leaves the machine. Watching only the spinner means the strongest privacy claim survives every feature we ship. The wait is the inventory; the work stays yours.