Browser extensions
Find out which AI technologies a web page loads, and how healthy its public trust surface is, from the browser you are using. Keryn for Chrome (and Edge), Keryn for Firefox and Keryn for Safari have the Keryn Web scanner in their popup, which scans a public GitHub repository you name, plus two scans of the page in the current tab:
- AI stack: which AI technologies the page loads (from a catalogue of more than 1,000 entries), as an AI-BOM with the basis
observed-in-browser, - Trust scan: the page’s public trust surface (security headers, policies, disclosures) as a scorecard.
Both scans request only the page’s own origin, as a visitor without cookies or credentials would.
Before you start
Section titled “Before you start”- The extension (a current Chrome, Edge or Firefox 140 or later), and a page to look at.
- A way to install it. None of the extensions is in a store yet: release 0.5.2 shipped the sites and the Keryn Cloud API only. Chrome Web Store, Edge Add-ons and addons.mozilla.org listings need accounts and review that are not done, and the Mac App Store listing for Safari is planned. Until then you load the built extension yourself (Chrome and Edge: load unpacked; Firefox: temporary add-on), and the install page is the authority on what is live.
- For the CLI step below, the Keryn CLI (install).
What each extension is allowed to do
Section titled “What each extension is allowed to do”| Extension | Id | Permissions it asks for | Status |
|---|---|---|---|
| Keryn for Chrome (and Edge, the same build) | store ids not assigned | activeTab, scripting, and https://codeload.github.com/* |
built, not published |
| Keryn for Firefox | keryn@keryn.dev |
the same; declares that it collects no data | built, not published; the Deep engine is not packaged in this build |
| Keryn for Safari | dev.keryn.safari (extension dev.keryn.safari.extension) |
the same manifest | built as an Xcode project; never run in Safari, so its behaviour there is unknown |
activeTab and scripting apply to the tab only after you click the toolbar icon; the extension has no always-on access to the pages you browse. The one fixed host is codeload.github.com, used to fetch the public repository you typed as owner/repo, with no cookies or credentials. The manifest declares no icons yet.
Save the page snapshot from the extension and feed it to the CLI with keryn scan --observed-browser snapshot.json; it joins the repository’s AI-BOM. Keryn Web takes the same snapshot under Settings, Import.
What you just proved
Section titled “What you just proved”A page snapshot is evidence of what a page loaded when you looked, saved as a file you hold. The extension requests only the page’s own origin, as an anonymous visitor would, and the snapshot joins the AI-BOM with the basis observed-in-browser, which is a different claim from “found in source” and is labelled so. Nothing about the visit is sent to Keryn or anyone else. The two requests an extension makes are the page’s own origin (for the page scans) and codeload.github.com (only when you scan a repository you named).
- Read the AI-BOM that results: read an AI-BOM.
- Scan the code behind the page: CI.
- What the extensions may call: local by design.