Support
Answers to the common questions — and a real mailbox when you need one.
Contact: support@llmmasker.com
We aim to reply within two business days. Please include your extension version (shown at the bottom of the Settings page).
How does masking work?
LLM Masker watches the prompt box on supported AI chat sites. When you paste or press send, it scans the text on your device and replaces anything sensitive — API keys, AWS credentials, JWTs, emails, phone numbers, cards, SSNs — with [REDACTED_BY_LLM_MASKER]. Then the prompt sends as normal. Nothing you type ever leaves your device.
Which sites are protected?
ChatGPT, Claude, Gemini, Copilot, DeepSeek, Perplexity and Google AI Mode (in Google Search). The extension runs only on those surfaces — and on google.com it touches only the AI Mode composer, never ordinary Search.
How do I choose what gets masked?
Right-click the toolbar icon → Options (or open Settings from the popup). Each detector has its own toggle, and the "When to mask" section controls whether scanning happens on paste and/or while typing. Masking at send time is always on while protection is enabled.
How do I pause it?
Click the toolbar icon and flip the switch in the popup header. That pauses masking on every tab instantly — no reload needed. Flip it back the same way.
Something wasn't masked. Why?
Detection is pattern-based. It catches well-formed keys, tokens and numbers, but it can't recognize free-form secrets (like a password written as a plain word) or values in unusual formats. Treat it as a strong safety net, not a guarantee — and if you find a real value that slipped through, tell us at the mailbox above; detector improvements ship regularly.
Something was masked that shouldn't be.
Turn off the relevant detector in Settings, or edit the placeholder out of the prompt before sending. If a normal string keeps getting caught, we'd genuinely like the (redacted!) shape of it — that's how the patterns get better.
How do I see or clear what's been masked?
The popup shows recent activity, and the dashboard has the full picture — counts, a 7-day chart, and a by-type breakdown. The log stores only the type and time of each mask, never the value. Settings → Clear activity data wipes it; the dashboard's Export JSON downloads it.
What does PRO add?
Custom rules, allowlists and a private audit log are on the way — see the PRO page for status.