Browser extension quickstart

After installing Sonomos, this five-step quickstart confirms the extension is working end-to-end — detection, the risk widget, and Cloak masking on a real AI platform.

It takes about two minutes.

Step-by-step

Open a supported AI platform. The cleanest test is claude.ai, gemini.google.com, or grok.com — these have full automatic Cloak support. Sign in if you’re not already.
Find the risk widget. Look in the bottom-right corner of the page for the Sonomos shield. It uses a traffic-light color:
- 🟢 Green — no PII detected on the page
- 🟡 Amber — low-to-medium sensitivity PII detected (e.g. a name, a generic email)
- 🔴 Red — high-sensitivity PII detected (e.g. SSN, credit card, MRN)
If you don’t see the widget, see Troubleshooting → Widget not appearing.
Trigger a detection with sample text. Paste this into the AI’s input box (don’t submit yet):
```
Hi, I'm John Doe (DOB 01/15/1990). My SSN is 123-45-6789
and you can email me at john@example.com.
```
The widget should turn red within a second or two and the detection count should jump. Click the widget to see exactly which entities Dagger picked up and at what severity.
Enable Cloak. Click the Sonomos widget on the page (or the toolbar icon) and toggle Cloak on. The widget border changes and the shield gains a small lock to indicate masking is active.
Send the message and watch Cloak work. With Cloak on, submit the message. Two things happen:
- The AI platform receives a sanitized version with the PII replaced by placeholders like [NAME], [SSN], [EMAIL].
- Your local view stays readable — Sonomos restores the original values in the conversation so your workflow isn’t disrupted.
Look at the AI’s reply: it should reference the placeholders, not your sample values. That confirms Cloak intercepted the message before it left your browser.

If all five steps worked, you’re set. Sonomos will keep watching every page from here on out.

What about other AI platforms?

Cloak does automatic masking on a small, focused list of platforms today. On every other site, Send Guard acts as a fallback — a banner or blocking modal that warns you before you submit text containing PII.

Platform	Protection level	Status
Claude.ai	Automatic masking	✅ Supported
Gemini	Automatic masking	✅ Supported
Grok	Automatic masking	✅ Supported
ChatGPT	Send Guard fallback	🟡 Warning only
Everywhere else	Send Guard fallback	🟡 Warning only

Recommended baseline configuration

For most users we recommend:

Cloak: on for every supported AI platform you use.
Send Guard: Banner + Modal (default) — banner for medium severity, modal for high severity.
Risk widget: visible — keep the traffic light in view so you build an intuition for what’s safe.
If you’re on Professional or Teams — sign in so detection history and compliance reports populate.

Troubleshooting the quickstart

Symptom	Likely fix
Widget never appears	Refresh the page after install; check for browser shields blocking extensions.
Cloak toggle has no effect	Make sure you’re on a supported platform (Claude.ai, Gemini, Grok). Other sites use Send Guard.
Detection count stays at 0	The page may have loaded before the extension; refresh and try again.
AI sees the unmasked text	Cloak may be off, or the page is unsupported. Check the widget — supported pages show a lock.

For more, see Common issues.

Next steps

Dagger detection engine Understand the detector categories, severity tiers, and how the AI models work on-device.

Cloak masking Configure Cloak for your workflow and see how interception works behind the scenes.

Using Sonomos with AI platforms Platform-by-platform guidance for Claude, Gemini, Grok, ChatGPT, and more.

Teams setup Roll Sonomos out across your organization with admin policy controls.