About AccessPulse

AccessPulse is built by an engineer who got tired of shipping inaccessible code without knowing — and even more tired of running axe-core on a page, seeing zero violations, and still finding alt text like alt="image" all over it.

The accessibility tooling landscape has a gap. On one end, there are free tools like axe-core, Lighthouse, and Pa11y. They work — and they're the same engines most paid scanners wrap underneath — but you need to wire up your own CI pipeline, build a dashboard, set up alerting, and maintain it all yourself. That takes 10-40 hours of DevOps work that most teams never prioritize.

On the other end, enterprise platforms like Deque, Siteimprove, and Level Access start at $15K/year and require sales calls. They are built for compliance teams at Fortune 500 companies, not for a developer who wants to know if their last deploy broke something.

AccessPulse fills the gap. It runs axe-core — the same engine Google, Microsoft, and the US government use — in a real browser on every URL you submit. You get a structured report with violations grouped by severity, the affected elements, and links to the relevant WCAG success criteria. Add AccessPulse/scan@v1 to your GitHub Action and you can fail builds that introduce regressions.

Then it does one more thing axe-core can't. For every image with an existing altattribute, a multimodal model looks at the image and judges whether the alt text actually describes what's on screen. Generic alt (alt="image", alt="photo"), filenames (alt="DSC_4823.jpg"), and descriptions that don't match the picture all pass axe-core because the attribute exists. They fail real screen reader users. AccessPulse flags them, explains why, and — for the broken ones — suggests a better alt as a starting point. For alt that's workable but could be sharper, we say so without writing it for you.

What this tool is and isn't. Automated testing catches roughly 57% of WCAG violations (per Deque's own research). The remaining 43% — alt text quality, reading order, cognitive load, video caption accuracy — requires human judgment or, in specific cases, machine learning. AccessPulse handles the automated layer reliably and is starting to chip into the other 43% with the alt text scoring above. The ML side is not perfect. We surface our findings as suggestions for a human to review, not auto-applied fixes. For full coverage, pair AccessPulse with periodic manual audits and, ideally, user testing with people who have disabilities.

AccessPulse is not an overlay widget. It does not inject JavaScript into your site. It does not claim to make your site "compliant." It tells you what's broken so you can fix it in your actual code.