ExtensionSafe reads browser add-on code before you install it, with reports tied to a package
Free queued checks and paid priority scans expose evidence that a permission list alone cannot supply.

A browser extension can request permission to read pages for a perfectly sensible reason. That permission doesn’t explain which pages it will read, what it will collect or where it will send the results. The permission dialog has the bedside manner of a blank check.
ExtensionSafe takes a store URL and downloads the published CRX or XPI package directly from Google or Mozilla. Its methodology describes calculating a SHA-256 checksum, reading the manifest, looking for relevant code paths and tracing information from its source to its destination.
The important distinction is between potential access and the use of that access. A tool that needs to modify a page may legitimately request broad permissions. A useful inspection has to follow what happens after the permission is granted, including remote instructions, form changes, cookie access and transmitted data.
The service combines deterministic checks with language-model review of the relevant code. Its documentation says it doesn’t execute the extension. That means it can produce a code-based account before the add-on runs in your browser, but it also sets a boundary around what was observed. A report about source code is different evidence from watching a running extension contact a server.
The methodology says findings cite filenames and source lines. That gives a reader something to investigate beyond a large reassuring score. Its own warning is blunt: “A high score is not a guarantee.” A hundred-point scale doesn’t make uncertainty disappear; it just gives uncertainty a rather tidy dashboard.
The current pricing page offers unlimited cached report views and fresh scans through a free queue. A five-dollar one-time pack buys five priority scans and twenty-five custom security questions. The distinction is queue priority, rather than a blanket charge for every new analysis; the credits are documented as non-expiring.
There is no account login. Paid credits are attached to a private key kept in the browser, and the site tells buyers to back it up. Someone with that key can spend those credits. Clearing browser data without saving it can therefore create an entirely separate problem from the extension being inspected (bonus administration).
Reports are public and shareable. Submitting a package for remote inspection should be understood on those terms, rather than confused with a private local test. For a custom or sensitive extension, its owner should decide whether that service model fits before handing over the material.
The methodology also supplies a correction route for maintainers who dispute a finding. That matters when automated interpretation can be wrong in either direction: the useful output is evidence another person can check, including evidence that needs fixing.
The score doesn’t get the keys.