PrivyPrompt - AI Prompt Privacy Guard

Before you paste a production key, customer row, private note, or internal document into an AI chat, PrivyPrompt gives you a local safety check. It scans prompt text and supported files inside your browser, shows a clear review panel, and lets you block, allow, or replace risky values with session-only aliases before anything is submitted. No account. No telemetry. No PrivyPrompt upload endpoint. Your prompt text, file text, OCR text, findings, custom terms, and alias mappings stay on your device. The destination AI website receives only what you approve. Use it when you: - Debugging with logs, `.env` files, API responses, tokens, or database URLs - Asking AI to summarize customer spreadsheets, support notes, contracts, or reports - Uploading supported documents, developer files, structured text, or clear screenshots - Working with client names, project codenames, employee/customer details, health/finance identifiers, or national IDs - Wanting local warnings for prompt-injection, jailbreak, system-prompt leakage, local-secret access, or secret-exfiltration wording Core features: - Scans prompts locally before submission on supported AI websites - Shows a review dialog only when sensitive or risky content is detected - Lets you allow, warn, protect, or block findings by category - Replaces selected values with session-only aliases such as `[PP_SECRET_1]` - Restores AI responses locally without inserting originals into the AI site's readable page DOM - Scans supported file attachments before upload, including documents, readable images, and text-like work files - Uses bundled local parsers and OCR assets with no remote code loading - Includes local presets for common privacy, healthcare, payment-card, and developer-secret workflows; these are shortcuts, not certification What it can catch: - API keys, access tokens, bearer tokens, JWTs, private keys, passwords, and database URLs - Email addresses, phone numbers, credit cards, IBANs, and network identifiers - National identifiers and country-specific ID formats with validation where reliable public rules are available - Healthcare identifiers, employee IDs, confidential business terms, and your own protected terms - Risky AI-tool instructions that try to reveal system prompts or exfiltrate local secrets PrivyPrompt is a privacy aid. It does not provide legal, regulatory, or security certification. Always review protected content before submission.