Prompt Firewall

Scans your prompts locally and warns before you send sensitive data to AI tools.

Use AI without leaking your life into it. Prompt Firewall runs in your browser and checks the text you are about to send from supported web-based AI chat pages. If it finds sensitive patterns, it shows a warning so you can cancel, redact in place, send anyway, or with Premium optionally request a safer rewrite using API credentials you provide. WHAT IT LOOKS FOR • API keys and tokens commonly found in developer configs and pasted code • Private keys, JWTs, and database connection strings • Payment and identity numbers where pattern rules apply • Emails, phone numbers, and postal-style addresses • Other secret and PII-style patterns (see the in-app rule list for details) PREMIUM • Custom watchlist, detector toggles, document modes, and policy templates • Optional Sanitise: a rewrite request is sent only from your browser to the HTTPS endpoint you configure, using your credentials — not proxied through Prompt Firewall servers. Anthropic’s browser API requires the header anthropic-dangerous-direct-browser-access: true; only your credentials are used. • Local audit metadata, exportable reports, and Privacy Mode to block outbound calls PRIVACY • Detection is local. There is no Prompt Firewall backend and no in-extension analytics SDKs. • Sanitise traffic goes only to the provider you configure. Audit entries store action metadata (site, action, score, categories), not your prompt text. • A small optional floating control on supported pages can be hidden in Settings. WHERE IT RUNS The extension activates only on the HTTPS origins listed under Permissions in this store listing. It does not claim affiliation with third-party services.