Early access · Berlin

Automated scans find 30% of the problems.
Agents find the rest.

Reviseberg runs the manual accessibility protocol — keyboard, screen reader, alt-text judgement, zoom, forms — as AI agents against your live site. Every finding comes with the transcript that produced it, so your team can check the work instead of repeating it.

No credit card. Report on your first 100 pages within one working day.

  • WCAG 2.2 AA and EN 301 549 mapped
  • Hosted in the EU, GDPR-compliant
  • Humans approve every change
82

Accessibility score

  • Since last crawl+6
  • Open issues41
  • Pages checked2,480
  • critical Checkout: focus lost after opening the payment dialog Keyboard agent
  • serious 19 product images described as “image123.jpg” Alt-text agent
  • moderate Error message not announced by the screen reader Screen-reader agent

Every finding carries the transcript that produced it.

Illustration of the Reviseberg dashboard. Values are examples.

The gap

A green scan is not an accessible website.

Rule-based scanners test what code can prove: contrast values, missing attributes, invalid markup. They cannot tell whether an alt text describes this picture, whether focus goes somewhere sensible after a dialog opens, or whether a screen reader announces the error a user just triggered. That is where most real barriers live — and where audits get expensive.

  • Scanners: fast, shallow

    Automated rules catch roughly a third of WCAG criteria. Everything requiring judgement stays untested — and unreported.

  • Manual audits: deep, rare

    An expert audit is thorough on the day it happens. Then you ship 40 releases and the report is out of date.

  • Overlays: risky

    Widgets that promise instant compliance do not fix the underlying code — and are named in a growing number of complaints.

How Reviseberg works

Agents do the testing. Your team does the deciding.

Eight specialised agents work through the manual protocol on your live pages and keep a full transcript of each run.

  1. Connect your site

    Add a URL, or connect staging with basic auth. No script on your site, no code change, nothing to install.

  2. Agents run the protocol

    Keyboard and focus, screen-reader output, alt-text relevance, zoom and reflow, forms and errors, media and captions, pointer and voice control, content sense.

  3. You get evidence, not a number

    Each finding shows the keystroke trail, the captured speech stream, or the image judgement that produced it — with the WCAG criterion it fails.

  4. Fixes get proposed, humans approve

    The AI drafts the fix — alt text, ARIA, a pull request, a content edit — and it waits in your decisions inbox. Nothing ships without a person.

One platform

Everything the compliance folder needs

Accessibility is the core. The rest of your digital quality comes along for the ride.

  • Accessibility

    WCAG 2.2 checklist, issue and page reports, prioritisation by effort and impact, PDF and document checks, guided EN 301 549 tests.

  • AI agent testing

    Human-equivalent coverage with full evidence, agreement rates against human auditors, and a “needs a human” queue for anything uncertain.

  • Compliance & reporting

    Progress per regulation — BFSG, EAA, EN 301 549, BITV 2.0, ADA Title II — plus a generated accessibility statement that stays current.

  • Documents

    PDFs checked for tags, reading order and machine readability, with bulk remediation and the pages that link to them.

  • Quality & SEO

    Broken links, misspellings, readability, page inventory, SEO issues and Core Web Vitals in the same crawl.

  • Fits your stack

    Jira, GitHub, Azure DevOps, your CMS, Slack. An MCP server lets your own AI tools read findings and propose fixes under your policy.

Why teams switch

Built for the release cadence you actually have

agents covering the manual protocol
8
of WCAG criteria a scanner alone can decide
~30%
keeps a reviewable transcript
Every run
hosting and data processing
EU

We are building Reviseberg with a small group of design partners — agencies, public-sector teams and shops facing the BFSG deadline. If you want your workflow shaping the product, this is the moment.

The Reviseberg teamBerlin

Deadlines

BFSG since June 2025. The complaints follow.

The German Barrierefreiheitsstärkungsgesetz (BFSG) implements the European Accessibility Act. Since 28 June 2025 it applies to consumer-facing online shops, banking, e-books, ticketing, telecoms and more. Market surveillance authorities act on complaints, and one complaint can end in a sales ban for the service.

Read the plain-language guide
  • Applies to B2C digital products and services in the EU
  • Micro-enterprises (<10 staff and ≤ €2m turnover) are largely exempt for services
  • The technical benchmark is EN 301 549, which points to WCAG 2.2 AA
  • You need an accessibility statement and a working feedback channel

Questions we get asked

Does Reviseberg make my site compliant automatically?

No — and be careful with anyone who says it does. We find barriers, prove them with evidence, propose concrete fixes and track progress. The fixes land in your code or CMS, approved by your team. Compliance is a legal assessment about your product, not a button in a dashboard.

Is this an overlay widget?

No. Nothing is injected into your site to “correct” it at runtime. We test what your users actually get and help you fix the source. An optional assistive widget exists for reading preferences, but it never counts as a fix.

How is this different from an automated scanner?

A scanner evaluates rules. Our agents perform tasks: they tab through the page and check where focus lands, read the page with a screen reader and compare the speech stream against expected semantics, and judge whether an alt text describes the specific image. Findings a scanner can only flag as “needs review” get decided — and when confidence is low, they go to a human queue instead of being asserted.

Where is my data processed?

In the EU. We store crawl results, findings and evidence for your sites; we do not sell data or use your content to train third-party models. Details are in the privacy policy, and a data processing agreement (AVV, Art. 28 GDPR) is part of the contract.

Do I need to install anything?

No. Public sites need a URL. For staging or gated areas, you can give us basic-auth credentials or allow-list our crawler.

What does it cost?

Plans start at €149 per month for a single site. See pricing — pilot customers get founding-customer terms.

See what an agent finds on your site

Send a URL. We scan up to 100 pages, run the agent protocol on your key journeys and send you the report — free, once, no strings.

Get a free scan Book a demo

Report within one working day.