How it works
Local detection
Regex, deterministic rules, and lightweight NER run entirely in your browser. No uploads, no API calls with your prompt text.
Reversible tokens
John Smith becomes <PERSON_1>, not [REDACTED]. Token mappings stay on your device so you can restore originals later.
AI-ready output
Copy a protected prompt and paste it into ChatGPT, Claude, or Gemini. The AI works with placeholders — your secrets stay local.
What we detect
Regex, deterministic rules, and lightweight NER — all running locally in your browser.
Protect ChatGPT / Claude / Gemini
Chrome extension with side panel + in-page 🛡 Shield button. Tokenize secrets locally, send via Shield API, sync Pro custom rules. Dev build: npm run build:extension.
Load unpacked: npm run build:extension → apps/extension/dist
Before vs after
Sending raw prompts
- Names, emails, and API keys go to AI providers
- No way to undo what was shared
- Compliance risk for client data
- Secrets may appear in training logs
With Privacy Guard
- Sensitive data stays on your device
- AI sees typed placeholders like <EMAIL_1>
- Reversible token mapping stored locally
- Zero upload — verify in DevTools
Common questions
No. Detection and tokenization run 100% in your browser. We never receive, store, or log your original prompt or token mappings.
Ready to protect your prompts?
Paste a prompt, detect secrets locally, and copy a safe version for AI chat.