PromptKavach – AI PII Filter

PromptKavach is an AI privacy firewall for Indian users. It scans what you type into AI chat interfaces in real time and intercepts sensitive personal information before it is sent — supporting ChatGPT, Claude, Gemini, and other popular AI platforms. Every detection runs locally in your browser. Your text never leaves your device for processing. There are no PromptKavach servers involved. ──── WHAT IT DETECTS ──── Indian Government IDs Aadhaar numbers (12-digit, with or without spaces) PAN numbers Passport numbers Financial Credit and debit card numbers (Visa, Mastercard, Amex, RuPay, JCB) Bank account numbers IFSC codes Identity & Contact Email addresses Indian mobile numbers (+91 and plain 10-digit) International phone numbers Dates of birth (when labelled) Credentials & Secrets API keys and secret tokens AWS access keys JWT tokens Inline passwords Network Public IP addresses Custom Keywords Add your own: project codenames, client names, internal terms ──── HOW IT WORKS ──── Manual review mode (default): When sensitive data is found, a review panel appears before the message is sent. You see exactly what was detected, choose which items to redact, and then send. Nothing is sent without your confirmation. Auto-redact mode: All detected items are silently replaced before sending. No overlay, no interruption. ──── WHY INDIA-FIRST ──── Every major competitor is built for Western data formats. None of them recognise Aadhaar, PAN, IFSC codes, or Indian phone number formats natively. PromptKavach was built specifically for Indian users handling Indian data. ──── PRIVACY GUARANTEE ──── All detection runs in your browser — zero external API calls No analytics, no telemetry, no usage tracking Open source — every line of code is publicly auditable on GitHub Minimal permissions — only the AI sites you use, nothing else See the full list of supported sites at github.com/nrbandi/guardprompt Open source. MIT licensed. Built in India.