Client-side prompt protection

Keep the raw prompt local. Copy only the masked version.

This layout keeps the process short: paste raw text, let the local masking pass rebuild the protected output, remove or regenerate masks if needed, then copy only the protected result into the LLM.

Everything stays client-side while you sanitize.

Masking runs in this browser session. The raw prompt stays local unless you copy text out of the page yourself.

Selected scope

🇺🇸 United States

Masking mode

Selected countries + global rules

Active matches

0 of 0

1. Set scope

Choose the country scope you want active, or keep the detector in global-only mode.

2. Paste raw version

Paste the original prompt on the left. No submit step is required in this layout.

3. Let it mask locally

Wait for the short local spinner while the protected output rebuilds on the right.

4. Refine and copy

Disable groups or single masks if needed, then copy only the protected output into the LLM.

Paste the original prompt here. The right side updates locally after a short spinner so you can inspect exactly what would be exposed.

This <output> block is the only text meant to leave the page. If anything sensitive still shows up, change the controls before copying.

Paste the raw prompt on the left and the protected version will appear here after the local masking pass.

Pick the masking scope, paste the raw prompt, and the protected output will rebuild locally.

Selected countries + global rules

Disable whole groups or individual masks below. Every checkbox updates the protected output immediately.

Credentials and API keys0 local matches

API keys, bearer tokens, passwords, and similar secrets should stay masked before you paste anything into a public LLM.

Mask this groupAlways keep masked