Every barrier in your component, with the WCAG 2.2 criterion it fails — and the same component again, fixed

Paste one component's markup — HTML, JSX, Vue or Svelte. Get the audit: each finding with its success criterion, its severity, who it blocks, the line and the fix. Then get the rewrite: same structure, same class names, barriers removed, the decisions only you can make marked TODO. Re-audit it with one button.

All three examples ship with a saved model run, so you can see the whole result — the findings, the manual test plan, the rewrite and its before/after scan — without signing in and without spending a credit.

The component's markup nothing pasted yet

Drop a file in, or choose one Everything is read in your browser. Nothing uploads until you run.

Framework — detected from the paste; override if it guessed wrong

What this is — optional, one or two sentences

Paste a component to price the run.

verdict

findings

blockers

passes

manual checks

scan covered

Findings

Blockers first. Each names the WCAG 2.2 success criterion it fails, who it blocks, the offending line, and the fix in your framework's spelling.

Keyboard map

Every interactive element the markup shows, the keys it must answer to, and whether the markup proves it does.

Element	Keys	Expected	Status

Manual test plan

What a static reader cannot settle — focus visibility, real contrast from classes, zoom reflow, screen reader announcements — with a concrete test for each.

Passes

Criteria the markup demonstrably satisfies. Only what is visible: a native button with text passes 4.1.2 for that control; a colour in a stylesheet nobody pasted proves nothing.

Scan reconciliation

changes

findings fixed

need a human

scan before

scan after

framework

The rewritten component

Same framework, same structure, same class names. Every TODO is a decision the markup could not make for you — they are listed below.

Before and after, by the free scan

The same mechanical scan, run on the original and on the rewrite. It proves the rewrite removed what it could measure — not that the component is now accessible.

Changes

Change	Criterion	Before	After	Fixes

Needs a human

Each is a TODO in the code above: what an image shows, what a label should say, which colour the brand allows. The page checked that every placeholder is really in the code.

What the markup cannot carry

Behaviour that lives in script — focus traps, Escape handlers, live-region updates, reduced motion — and where in this component it belongs.

Test script

Run it with a keyboard and a screen reader over the rewritten component before you call it done.

#	Step	Expect

Raw model reply

The reply did not parse as the structured result, so it is shown as it arrived.

What this does, and what it does not

The scan is a reader of markup, not a browser. It builds the element tree from your paste — tolerating JSX, Vue, Svelte and Angular spellings — and checks what a static reader can prove: an image with no alt, a field with no label, a control with no name, a div wearing a click handler, a positive tabindex, an ARIA reference to an id that is not there, a heading that jumps two levels, a dialog with no name, a table with no headers, radios with no fieldset, autoplay, zoom blocked, a focus outline removed. Where both a text colour and a background are literal — in an inline style or a <style> rule — it computes the WCAG contrast ratio and holds it to 4.5:1, or 3:1 for large text. Everything it finds goes to the model as facts it must reconcile, and the reply is checked against them afterwards — if the model skips one, this page says so.

The model finds what the scan cannot: meaning carried by colour alone, a visual-only asterisk, a form that never says what went wrong, a login without autocomplete, focus that will go nowhere when a dialog closes. It cites the WCAG 2.2 success criterion for each, sets a severity that is about the user rather than the code, and derives the verdict by a fixed rule the page re-checks. The rewrite keeps your framework, your structure and your class names, prefers native elements over ARIA, and marks every decision it could not make with a TODO the page confirms is really in the code. Then the free scan runs again on the rewrite, so you can see what it removed — and a button hands the rewrite back to the audit lane.

Nothing leaves this page until you press the run button. The page reads your markup in the browser, sends it to the model only when you press that button, and hands you back text. A token or password in the paste is flagged for rotation and never repeated. A verdict of "ready for manual testing" is exactly that: the static review found nothing left, and the manual plan is what stands between the component and a claim. Nothing to hand? Load the , the , or the . All three replay a saved run for free.