AI Privacy Cleaner
A local-first tool that scrubs sensitive data out of text before you paste it into an AI — names, emails, keys, tokens, and more are detected and consistently replaced, entirely in your browser with nothing sent to a server.
People paste real work into ChatGPT and other AI tools all day — support tickets, logs, code — and that text is full of things that shouldn't leave the building: customer names and emails, internal URLs and IPs, and, worst of all, live secrets like API keys, database URLs, and auth tokens pasted straight from code. Most 'redaction' tools make it worse: they send your text to a server to scan it (defeating the whole point), or they blank out matches inconsistently, so the same name becomes a different placeholder each time and the text stops making sense to the AI.
AI Privacy Cleaner is a redaction tool that runs entirely in the browser — no backend, no database, no account, and no AI API call — so the sensitive text never leaves the machine. Detection engine: a library of purpose-built matchers finds names and company-like names, emails, phone numbers, URLs, domains and IPs, addresses, ticket/order IDs and long identifiers, and — critically for developers — API keys, bearer tokens, JWTs, private keys, database URLs, and passwords in code (matching real formats like sk-, ghp_, AKIA, xoxb, and PEM private-key blocks). Scan modes: general text, a stricter scan, and a code/log-focused mode that switches on the secret-detectors. Consistent replacement: the same value maps to the same placeholder throughout a document, so the cleaned text stays coherent and the AI can still reason about it. Built as a fast static site in vanilla HTML, CSS, and JavaScript (no framework), with live match/category stats and a light/dark theme, and shipped with a Docker + Nginx setup for deployment on Coolify.
A finished, deployed tool that addresses a real everyday privacy risk — pasting secrets and PII into AI — with a genuinely private, client-side design. Small in footprint, sharp in purpose.