Chrome Extension · On-device AI · Privacy & Security
Your calm guardian for safe browsing
Scams, phishing, and data leaks — guarded, privately. Sentinel does the technical
checks a person can't, right in your browser, and warns you in plain language. It
runs entirely on-device with Chrome's built-in AI. Nothing about your browsing ever
leaves your computer.
Warns — never blocks. Never nags, never says a site is "safe."
Guards your browser in real time Runs on your device Plain-language warnings You stay in control
Scam Radar
Spots fake and phishing sites your eye can't catch
Sentinel checks every site for lookalike domains and scam tactics — combining
code-level facts (lookalike hostnames, HTTPS, known-bad lists) with the on-device
AI's read of the page's wording.
Detects lookalike sites before you click a link
Flags likely scam & phishing pages with a calm, clear warning
Escalates to an ask the moment a password or code is at risk
Works even without AI — the code checks always run
The core four
One engine, many lenses
AI judges language and intent; code handles the facts. Together they guard the moment of harm — in your browser.
Never-Share GuardCodes & passwords
Warns before a one-time code, 2FA code, CVV or recovery code goes somewhere it shouldn't — a chat, an AI tool, or an unknown site. Stays silent in the real code field.
Scam RadarFake & phishing pages
Lookalike-domain, known-scam and HTTPS checks combined with the AI's read of the page language — escalating at the credential moment.
Background WatchSilent data sends
Watches what the page sends and pauses before your data goes to an unknown or lookalike domain — Allow once, Block, or Always allow.
Agreement GuardFine print
Summarizes cookie / terms / permission pop-ups in one plain sentence — and flags camera, mic, location and clipboard requests.
Advanced protections
Six more guards, quietly watching
Redactor
Offers to mask a big block of personal text — emails, phones, cards — before you paste it into an AI chat.
Lookalike-link warning
A quiet glow when you hover a link to a lookalike domain — and a full stop before you click a known-scam link.
Dark-pattern flag
Names the trick when a page uses pre-checked opt-in boxes or a fake countdown timer to rush you.
Wrong-site password
Uses on-device memory to catch a password you've used elsewhere being reused on a lookalike or known-scam site.
Sneaky download alert
Pauses a download the page started that you didn't click — the file hasn't started until you say so.
Insecure-page warning
Flags a non-HTTPS page that's asking for a password or card — folded right into Scam Radar.
Never-Share Guard
Protects your OTPs and passwords
Most scams — including deepfake calls — converge on getting a code out of you.
Sentinel warns the moment a code or password is about to go somewhere it shouldn't,
even if someone on a call is asking for it.
Catches risky sharing the moment it happens
Stays silent when you enter a code into the real one-time-code field
Reads "send me your OTP" messages in English and Hinglish
You're always in control — we never send anything
Private by design
Your data never leaves your device
All protection runs on-device. Nothing about your browsing is sent to us — there
are no "is this site safe?" lookups, because those would transmit where you go.
Data flows in only: Sentinel downloads public threat lists and
matches them locally.
No cloud, no servers — 100% on-device
No analytics, no telemetry, no tracking, no accounts
Pattern memory stores a salted hash of a password — never the password
Built on trust — transparent and honest about what it does
Quiet by design
It interrupts only when it truly matters
Severity = how sure something's wrong × how bad it'd be. That maps to three calm levels — most of the time you see nothing.
1
Silent
Normal, expected, low-risk pages. Nothing shown — the shield just keeps a quiet grey heartbeat while it checks.
2
Quiet glow
Most warnings — sketchy pages, agreements, oversharing. The shield turns orange with a dot; details only if you click.
3
Ask
Only real, hard-to-undo harm — a secret about to leak, or a suspicious background send. A panel opens and asks you.
Agreement Guard
Understand what you're agreeing to
Plain-language summaries before you click Accept. Sentinel reads cookie and terms
dialogs locally and tells you, in one sentence, what you'd actually be granting.
Explains cookie banners in plain language
Highlights what gets shared and why it matters
You decide with clarity and control
Honest about our role — we warn, never guarantee
🛡️
Honest about scope — false confidence is the worst outcome.
Sentinel guards what happens inside your browser. It does not protect your operating system, other apps, phone or video calls, SMS, or your network — and it will never claim to. For those, use real OS-level security: antivirus, OS updates, 2FA and a password manager. Sentinel does the technical browser checks a person can't, and tells you honestly where it can't help.
Before you install
What you'll need
Sentinel's AI features use Chrome's built-in Gemini Nano. Without it, the code-based checks still run.
Chrome 138+
Desktop Chrome (not mobile). On 128–137, enable the Prompt API flags and restart.
~22 GB
Free disk for the one-time on-device AI model download.
GPU / RAM
A modern GPU with >4 GB VRAM for the local model.
Free
No accounts, no API keys, no subscription. Ever.
FAQ
Questions, answered
No. Sentinel warns and asks — it never hard-blocks. Every warning offers a clear choice (continue anyway, leave, allow once). It also never tells you a site is "safe."
Yes. Everything runs on-device. Sentinel never sends the URLs you visit or anything you type. Its only network activity is downloading public threat lists, which contain no information about you — see the privacy policy.
No — and it will never claim to. Sentinel guards your browser. For your OS, calls, SMS and files, use real OS-level security. Honest scope is the point.
Sentinel still runs all the code-only checks — lookalike domains, HTTPS, regex secrets and the network watch. The popup just notes that the AI language features need Chrome desktop with built-in AI.
No. It's silent by default and never warns twice for the same thing. You can "always allow" a site, and a Quiet mode snoozes glow-level warnings for an hour.
Browse with a calm guardian
Install Sentinel and let on-device AI watch for scams, phishing and data leaks — quietly, and privately.
Add to Chrome
