Ghost Mode

Nothing is stored

With a Ghost key the bodies of requests and answers are not saved. What stays is the token counter needed for billing, not your text.

Turned on per key

A Ghost key for sensitive work, an ordinary one for everything else. Same endpoint, same models.

Verifiable, not a promise

The answer carries a signed no-retention receipt, and the API tells you what exactly is kept for your key.

How it works

Create a key and turn Ghost Mode on for it.
Use it anywhere: the API directly, Codex, Cursor, any OpenAI-compatible client.
A request made with that key skips storage. Only the token counter is recorded.

Honest about the limits

The request still goes to the model vendor — otherwise there is no answer. Ghost Mode is about what we keep: nothing but the token counter. There is no claim of total privacy here; there is zero retention on our side and a way to check it.