Skip to content

Protect a request with ZKR.

ZKR masks detected sensitive values before supported cloud requests and restores matching values in the reply. Your task remains readable while fewer personal details are shared.

Protect text

  1. Choose a supported cloud route.
  2. Turn on ZKR near the chat input.
  3. Write your prompt. Highlights and the control’s counter show detected values.
  4. Review the prompt and send when ready.

Your original prompt stays editable. If you turn ZKR off, it stays as you wrote it. The control does not permanently replace your draft with the masked version.

Try it with fictional details

Draft a short email to Jamie Park at jamie.park@example.com confirming that the receipt total is $42.80. Keep it friendly and ask for confirmation by Friday.

Watch the highlights and counter as you type. Check which values are detected before sending. The amount and deadline give the model something useful to work with; the sample contact details are fictional.

Protect a file

Supported documents and receipt images need a review step. Prepare the protected copy, check the extracted text and masks, then send. Follow the file review guide.

Read the result

Matching values can be restored when the answer arrives. The request’s receipt describes the protection applied. Understand receipts.

Know the limits

Detection can miss personal details or mask text you intended to keep. Check what the task contains. ZKR is not proof of complete anonymity, an encrypted storage system or a guarantee about remote processing.

When a task requires every detail to stay on your hardware, use an appropriate local model and consider whether any online tools are enabled.

A private AI workspace, built to put you in control.