the fine print 15 questions no account to read any of it
Answered at length rather than in a sentence each. Where an answer turns on a number or a limit, the number is the part underlined. The full disclosure of what is read and what is sent is on the privacy page; the agreement itself is in the terms.
100 heuristics across six weighted categories - markup, CSS, runtime, copy, accessibility and page weight. Each awards partial credit against a budget and quotes the value it measured, so a score of 62 comes with the sentence that produced it. The scoring runs on the server, not in the extension, so it cannot be read off or edited from the page it is judging.
No. It means the markup carries the fingerprints of a generator: repeated utility strings, inline SVG flooding, unstyled component defaults, buzzword headings. Plenty of good sites are generated and plenty of hand-written ones are a mess. The report is a starting point for a refactor, not a verdict on the work.
DevTools answers one element and one question at a time. This reads the whole page at once and hands back a ranked list - every element that overflows the viewport, every stacking context, every image shipped at twice the size it renders, a heatmap of which rule actually won. It also answers questions DevTools never asks: what the page was built with and how each detection was made, what its design tokens are, and how much of the markup reads as generated. It floats over the page rather than docking beside it, so what you are inspecting keeps the width it had.
So did we. Seventy-five tools and diagnostics across nine surfaces in one panel is one install, one shortcut and one place to look - and, more to the point, one permission list. Four extensions that each read every site you visit are a wider surface than one that reads the tab you are looking at when you click it. Compare the manifests rather than the feature tables; that is where the difference is.
Free is for looking, and looking is most of it: the inspector, the extractors, the stack detection with the evidence for each match, the SEO and accessibility readers, and the tools and diagnostics that answer from the document already in front of you. Pro is for taking something away - a generated theme or token file, a client-ready report, a WCAG conformance statement, the third-party register - and for the readings that cost a request, the audit past the free three among them. Every locked feature is marked where it sits, with a sentence saying what it does once unlocked, so you meet the line while reading rather than at the checkout.
One payment, no renewal, and it carries exactly what the monthly plan carries. That is a constraint in the code rather than a promise in the copy: both plans are held to one shared feature list, so neither can quietly grow a perk the other lacks. The prices on this site are the prices Stripe charges; if you ever find the two disagreeing, the checkout page is the one that is right and the other is a bug we want reported.
14 days from the payment. Email and ask - there is no form to fill in and no retention flow to sit through. Fourteen rather than thirty because it is the withdrawal period EU consumer law already gives you for digital content bought online, so the promise and the obligation are one thing instead of two. Payment runs through Stripe and we never see a card number.
That is what the changelog is for. The release list on this site and the What’s new screen in the panel render the same array, so the rate is visible rather than promised. Adding a tool to the panel is a row in a table by design, which is why the count keeps moving.
Most of it does not depend on us. Everything that reads the page in front of you runs in the browser and calls nothing - the inspector, the extractors, the stack detection, the SEO and accessibility readers, and every tool and diagnostic that works on the open document. What would stop is the handful of readings that leave this machine: the audit, the Server tab’s probes, field data from the Chrome UX Report, the advisories against a detected version, the sitemap crawl, and anything published at a link. The panel would keep working; it would answer fewer questions.
Yes, personal and commercial both. The client report is built for exactly that - a branded export of a finding at a link that stops working on a date you choose, instead of a screenshot in an email.
Yes. It reads whichever tab you are on when you click, which includes localhost, a staging host and anything you are signed in to. There is no list of sites it works on because there is no list of sites at all.
A hostname, integer counts, class-name strings, and the page title plus heading text after truncation and redaction. No body text, no input values, no full URL. The privacy page lists every field.
An audit costs a server request; nothing else in the extension does. The inspector, extractor and the three playground tools run entirely in your browser and are free without limit. A device id is minted locally on first run - reading a page asks for no account and no email address.
activeTab and scripting, granted for the tab you are on when you click. That is what lets it read the page you asked about. It does not request access to all sites, and neither built manifest contains a host permission.
Chrome, Edge and any other Chromium browser, and Firefox from version 140. It is the same panel on both - one build target each, and the two manifests are structurally identical. The Firefox floor is 140 because that is the version with the built-in data-consent screen; below it an add-on has to ship its own, and a second parallel consent path for the one security-relevant statement this product makes is worse than a higher floor. Safari is not shipped.
still unanswered? write to support