Prompt Armour — Redact Secrets & PII from ChatGPT, Claude & Gemini

Automatically catches passwords, API keys, and personal data in your prompts and redacts them before they reach any AI chatbot.

Once you paste sensitive data into an AI chatbot, you can't take it back. Prompt Armour catches it before it's sent — automatically, locally, in your browser. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ HOW IT WORKS 1. Type or paste into any supported AI chatbot 2. Sensitive data is highlighted in real-time as you type 3. Click Redact — secrets are replaced with safe placeholders 4. Send your prompt. The AI gets context without your secrets. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT IT CATCHES Developer Secrets: - AWS Access Keys, Secret Keys, ARNs, EC2 instance IDs - Database connection strings (MongoDB, PostgreSQL, MySQL, Redis) - API keys, Bearer tokens, Authorization headers, session cookies Personal Information: - Passwords, emails, phone numbers, SSNs, credit card numbers Network Data: - IPv6 addresses, MAC addresses, internal network IPs Unknown Secrets: - Shannon entropy scanner catches high-randomness strings that look like passwords or keys — even if they match no known pattern Names & Locations: - Lightweight NLP detection for names and places in your text ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ REDACTION MODES → Smart Tags: [EMAIL-1], [AWS-KEY-2] — keeps context for the AI → Masking: ********** — classic obfuscation → Full Redaction: [REDACTED] — clean removal ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WORKS ON - ChatGPT (chat.openai.com) - Claude (claude.ai) - Google Gemini (gemini.google.com) UPCOMING ON 28 APR 2026 - Microsoft Copilot (copilot.microsoft.com) - Perplexity (perplexity.ai) - Grok (grok.com) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ BUILT FOR PRIVACY - 100% client-side — all detection runs in YOUR browser - Zero data transmission — nothing sent to any server, ever - No accounts required — install and it works immediately - No analytics, no tracking, no telemetry - Shadow DOM isolation — our UI never touches the host page We built a privacy tool. It would be embarrassing if it wasn't private. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ DESIGNED FOR CONTROL - Toggle individual detection categories on or off - Undo any redaction with one click - Hover over redactions to selectively restore just that item - Whitelist false positives for the current session - Granular per-category settings — block AWS keys but allow emails, or any combination you need ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHY THIS EXISTS Samsung engineers leaked proprietary source code through ChatGPT. Amazon warned all employees about sharing data with AI tools. Italy temporarily banned ChatGPT over privacy concerns. You can't control what AI companies do with your data. You CAN control what reaches them in the first place. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FEEDBACK & SUPPORT Independent and bootstrapped — built by a developer who got tired of watching people accidentally paste secrets into AI. Actively improving based on user feedback. Found a bug or have a feature idea? We want to hear it. Website: https://prompt-armour.vercel.app Email: sharma.apoorva1@outlook.com A 5-star review helps more than you know.