Not one scanner — a department
Meet the team that audits your site.
When you hit Analyze, you're not running a script — you're commissioning a department. Nine specialists, each with one job, passing work to each other the way a good agency does. Every desk below is a real component running on our fleet today.
The specialists
🕸The CrawlerReads your pages' actual HTML, source-level — the fast first pass that maps every form, image and control.hands findings → the Findings Deskheuristic engine, server-side
🖥The Browser TesterOpens your site in a real browser and runs an industry-standard rule set, the way a visitor's browser sees it. Rolling out: until it is live for your account, your report runs on the Crawler and the Second Opinion.hands violations + WCAG tags → the Findings Deskreal browser · rolling out
🔍The Second OpinionAn independent auditor from outside our shop — WebAIM's WAVE engine cross-checks us, because we don't grade our own homework alone.confirms / challenges findingswave.webaim.org API
⚡The Performance AnalystMeasures Core Web Vitals while the browser is open — slow, shifting pages hurt disabled users first.adds the performance band to your reportLighthouse · CWV
📚The LibrarianKeeps the rulebook: every WCAG criterion, every law, every detection method — and attaches the right citation to every finding.stamps each finding with rule + lawaccessibility KB · RAG index
🧭The ResearcherCloviScout keeps the Librarian's shelves current — new rulings, new WCAG drafts, competitor claims worth fact-checking.refreshes the KB continuouslyCloviScout research agent
🧠The Pattern KeeperThe learning agent. Remembers 8,969 findings across real sites and which fixes actually held up — so noise gets filtered and real risks surface first.re-ranks findings by real-world prevalencelearned patterns · flywheel
✏️The Fix EditorDrafts the exact code change for each finding — old line struck out, new line ready — never applied without sign-off.sends diffs → your Fix QueueAI via fleet LLM gateway
⚖️The Legal ClerkMaps your findings to litigation exposure and drafts your accessibility statement and proof-of-effort file.writes the legal section of your reportCloviLegal
🫵You — the ReviewerThe only one with authority to ship. Approve, edit, or reject every fix. Your decisions are the department's training signal.every approval teaches the Pattern Keeperthe fix queue
How your report gets made — the hand-offs
- Independent auditors examine your page. The Crawler reads the source and the Second Opinion audits independently; the Browser Tester joins as it rolls out.Crawler + Second Opinion today, Browser Tester rolling out
- Findings meet at the desk. Duplicates merged, disagreements flagged, severity weighed — forms first, because they cost you customers before they cost you lawyers.Findings Desk
- Every finding gets its citation. The exact WCAG criterion and the law behind it — so your developer, or a court, can verify every line.The Librarian
- Experience filters noise. The Pattern Keeper checks each finding against what's proven real across thousands of prior findings — and demotes what usually turns out to be a false alarm.The Pattern Keeper
- Fixes get drafted, not applied. The Fix Editor writes the before/after diff for each issue and queues it for you.The Fix Editor → your Fix Queue
- The legal picture is written. Exposure overview, conformance matrix, remediation plan in priority order.The Legal Clerk
- You decide. Approve one, bulk-approve twelve, reject anything. Nothing touches your site without you.You
- The report is bound — and the department gets smarter. Score, POUR pass rates, AA matrix, remediation plan, Core Web Vitals. Your decisions flow back to the Pattern Keeper for next time.bound report → learning loop
See a finished full report — every section above, on a real site →
3independent audit engines per scan
8,969findings in the Pattern Keeper's memory
595scan runs feeding the flywheel
emergingthe department's self-graded confidence tier
The honest part: the department publishes its own report card and it starts at the bottom — a fix pattern only graduates when it holds up on multiple independent sites, and human-judgment items are never auto-passed. Ten desks, one rule: nothing fake ships.