Free scanFree scan

Website accessibility testing platform Find the issue.
Know the next step.

Reviseberg combines rule-based scanning, an agent that presses real keys, recorded manual results and AI suggestions — with a person deciding every one of them, because nothing here changes your site.

Get a free scan Book a demo
82

Accessibility score

  • Score since last crawl+6 points
  • Open issues41
  • Pages checked2,480
  • critical Checkout: focus lost after opening the payment dialog Keyboard agent
  • serious 19 product images with no text alternative axe-core rule
  • moderate Filter buttons that no Tab key reaches Keyboard agent

Every keyboard finding carries the keystroke trail that produced it.

Illustration of the Reviseberg dashboard. Values are examples.

Find

Testing that says what it could not decide

Automated rules and checks of our own run on every page of every crawl. The keyboard agent walks the journey you name. Nothing they find is a pass: a criterion stays partial or untested for a person to settle and record.

  • WCAG 2.2 A and AA per page and per issue, the criteria no check reaches marked untested and the rest partial until a person decides
  • EN 301 549 clause by clause, from the same criteria, for the procurement questionnaire
  • Keyboard and focus: order, traps and visible focus, with the keystroke trail and a screenshot at the failure
  • Every axe-core rule at two viewport widths, contrast measured on rendered text
  • Checks of our own: every page at 320 px, zoomed to twice the size and with wider text spacing, contrast against the pixels behind text and in hover and focus states, captions, motion and time limits; from an account, controls, hover content and forms with test data on one page per template
  • Manual test results with author, method and a twelve-month expiry
  • Quality and SEO: broken links, reading level, page inventory, the SEO basics, Core Web Vitals lab metrics, cookies and third-party requests
  • Documents: which pages link to PDFs and Office files, and a structural check of your own PDFs — tags, language, a displayed title, figure alternative text
  • Not included: a full PDF/UA check. Office files are listed, not opened.

Prioritise

A ranked list, not a 900-row export

Findings are grouped by rule and ranked by what fixing them gives back, so “label the search field” with its forty pages replaces forty rows. Each issue shows the pages affected, the criteria it clears and how much of the score is on the table.

  • Points to gain: what moves the score most per hour of work
  • Page groups by URL pattern, and site comparison for multi-brand or multi-country setups
  • Goals and targets measured by the crawl, never ticked off by hand
  • Assignment, priority and notes per issue — and no “done”: an issue is fixed when the crawl stops finding it

Fix

AI proposes.
A person decides.
The fix is yours to make.

Where an answer can be checked, AI drafts one: alt text for an image, a corrected snippet for a rule, a plain-language rewrite, a translation. Each suggestion is marked against the rules it was given and waits for a person. Accepting records a decision; the change in your source is yours to make. If you run the optional visitor toolbar, an Admin can let it show accepted alt text, labels, titles and languages as temporary fixes — off by default, labelled, and the issue stays open in every scan until your source has the fix.

  • Alt text written from the image and the page around it, checked against the guidance a reviewer uses
  • Corrected snippets for the rules a snippet can settle
  • Plain-language rewrites scored with the readability measure the crawl uses
  • Your coding agent reads accepted suggestions through the MCP server
  • Not included: contrast fixes. That is a decision about a palette, and a snippet cannot settle it.

Prove

Documentation that survives an authority request

Coverage per regulation, an accessibility statement drafted from the evidence, scheduled reports for management, and the trail behind every claim.

  • EAA and BFSG, BITV 2.0 and ADA Title II views on the same evidence
  • An accessibility statement drafted from the crawl and the recorded manual results, versioned, published at a public address
  • Scheduled reports by email, weekly or monthly, to people who need no account
  • Monitoring: crawl history, uptime checks and an alert when a score drops after a release

Integrations

It plugs into the tools you already run

Chat alerts, webhooks, an API for your own scripts and a pre-publish check for CI. Plus an MCP server, so your own AI assistants can read findings and accepted suggestions. Jira, GitHub and CMS connectors are not built yet, and the settings screen says so rather than showing a button that does nothing.

  • Slack
  • Microsoft Teams
  • Webhooks
  • REST API
  • MCP server
  • CI pre-publish check
  • CSV export

Further reading

  • Visitor toolbar

    One script tag that lets each visitor change text, colours, motion and reading aids for themselves — without changing your site, and without claiming to.

  • Methodology

    How Reviseberg tests: axe-core rules, a keyboard agent, a status for every WCAG criterion, how the score is calculated, and where automation stops.

  • MCP server

    The Reviseberg MCP server puts WCAG findings, conformance status and alt-text suggestions into Claude Code, Cursor, VS Code and claude.ai.

  • API & CI

    A REST API, webhooks and a pre-publish check for GitHub Actions and GitLab CI that fails the build when a page breaks your accessibility threshold.

  • Keyboard agent

    Keyboard accessibility testing with real keys: the agent tabs through your pages, finds keyboard traps and invisible focus, and records every step.

Start with your own site

A free scan is the fastest way to see what the rules and the agent find on your site that your current tooling does not.

Get a free scan See pricing