Help & support.

No. Your prompt text never leaves your browser, and the scanner makes no network requests. The only request the extension ever makes is an optional Pro license check, which sends your license key, never your prompts. The full breakdown is in the Privacy Policy.

Detection, the pre-send warning, element redaction, and the sidepanel Redactor are free forever — plus up to 5 custom patterns and 3 custom AI sites. The text, area, and auto-redact tools each give you 3 free previews to try them; after that they're part of Pro. Pro ($3.99/mo or $37.24/yr) also lifts the pattern and site caps and adds the stats dashboard, pattern import, coded redaction, per-site detection rules, and custom redaction labels you edit right on the page (more on those below).

There's no trial and no account. The free previews don't reset on reload — they're a real taste of Pro, not a timer. When you're ready, a license key unlocks everything; see Pricing.

Not with the default labels — every email becomes the same [email redacted], so the model can't tell two people apart. If that matters, Pro adds a Coded redaction style (gear menu → Settings → Redaction style). It replaces each value with a consistent numbered token instead: [EMAIL_1], [EMAIL_2], [PHONE_1]. The same value always maps to the same token, so the AI keeps the relationships while your real data never leaves. The mapping is in memory only and is never saved.

Yes, on Pro. Hover any flagged value on the page, type exactly what it should become — [client], a ticket code, whatever fits — and redact. Your label is checked so it can't itself be something flaggable, and it sticks to that value for the rest of your session, so every other place it appears redacts the same way. It's all in memory; nothing is stored.

Yes, on Pro. The Detection card follows whichever AI site is in the tab in front of you. Hit Customize and any category you change applies to that site alone — every other site keeps your default, and two tabs on different sites run their own rules at once. Reset sends a site back to your default, and you can turn Redactopia off for a single site without disabling it everywhere. Handy when you want everything strict on a work tool but lighter on a personal chat.

Only if you turn it on. There's a Sync toggle in the gear menu, off by default. When it's on, Redactopia mirrors just your custom patterns, your custom AI site list, and your detection preferences to your other devices through your browser's own encrypted sync — your Google or Microsoft account, never a Redactopia server. Your prompts, findings, redacted text, and Visual Redactor records never sync. Leave it off and everything stays on this device. (Site permissions don't sync, so a synced custom AI site needs a one-time permission grant on each new device.)

By default, Redactopia covers a handful of popular AI chat sites; that's where prompt scanning and the warning modal show up automatically. The exact list depends on your browser build, since one of the sites needs the Chrome build for platform reasons (more on that below).

For anything else, the Custom AI sites surface lets you add other AI chats yourself. Open the side panel → Custom AI sites → type a hostname (like chat.mistral.ai) → Add. If the Visual Redactor is on, the broad all-sites permission already covers the new site and no extra prompt appears. If it is off, your browser asks for that site specifically. Once added, the same scanner and warning modal apply.

Side panel → Custom AI sites → type the hostname → Add. If the Visual Redactor is on, no permission prompt appears; the broad all-sites permission already covers it. If it is off, your browser asks for that specific site. Once added, scanning starts and the warning modal blocks sends that contain sensitive content.

If the warning fires on the wrong buttons (Search, attach, model picker, etc.), click the ✦ button on the site's row and pin the actual send button. The over-trapping stops after that. And if a site you calibrated later changes its layout, Redactopia tells you the calibration went stale — the on-page counter gets louder and a ⚠ appears on that site's row — so you can re-pin instead of being caught out.

Copilot is a Microsoft product, and Edge, also Microsoft, doesn't allow third-party extensions to run on Microsoft's own pages. That's a platform rule that applies to all extensions equally, not something Redactopia-specific. If you want Copilot coverage, use Chrome. Everything else works the same in both builds.

Only if you've allowed it. Browsers disable extensions in private windows by default. Open your browser's extensions page (chrome://extensions or edge://extensions), find Redactopia, and turn on the "Allow in InPrivate" / "Allow in incognito" option. Until then, the extension is inactive in private windows.

Yes. Open the side panel. When you're on a supported AI site, the status bar at the top shows a per-site toggle. Flip it off and protection pauses for that site. The Detection card lets you turn individual categories on and off too. Settings save automatically.

No. The scanner runs on a debounced check, not on every keystroke, so typing stays smooth in normal use.

Yes. Side panel → Custom Patterns. Add a name, a JavaScript regex, and a severity. Patterns are validated before save and run alongside the built-in detectors. There's also a Browse library button with a growing set of starter patterns you can add with one click — useful for common shapes like internal ticket IDs, employee badges, project codes.

Chrome and Edge are both first-class. Use whichever you already use. The two builds are nearly identical — if you're wondering about Copilot specifically, see the question above.

Firefox isn't supported yet. It uses a different extension API than Chrome and Edge, so it needs its own build — which we haven't shipped. It's on the roadmap, with no firm date yet.

Yes. AI sites change their interfaces often, so we sweep the send-button detection across supported sites when it breaks, add new detection categories, and ship security hardening. Your extension store listing shows a "Last updated" date, so you can always see when the latest build went out.

Yes. Open the side panel → gear menu (Settings) → turn off Highlight detections in the text. That hides the colored outlines drawn around sensitive text as you type. The count pill in the corner of the composer and the send-time check both stay on, so you still get a heads-up — you just lose the inline outlines. The warning modal works exactly the same either way.

The badge count shows how many sensitive findings are in the current input. An amber badge means at least one high-risk finding (API key, password, SSN, etc.). A blue badge means standard-risk only (email, phone, etc.). The badge clears when you dismiss the warning modal or navigate away.

Open the side panel → gear icon (bottom-right) → Settings popover. The Your usage section shows total scans run, total findings caught, and when you first started using it. All counts are local — they never leave your browser.