WCAG MCP Server

Ask any LLM to cite a WCAG success criterion and there's a meaningful chance it will hallucinate the number. It will confidently say "this fails 1.4.3" when the issue is really 1.4.11, or invent a 4.1.4 that doesn't exist, or pull a quote from "Understanding WCAG" that nobody at the W3C ever wrote. wcag-mcp closes that gap by giving the AI direct, structured access to the actual WCAG 2.2 documentation — guidelines, success criteria, techniques, and the Understanding documents — so when your assistant says "this fails 1.4.3 contrast," it's pulling from the spec, not from a half-remembered Stack Overflow answer.

Built by joe-watkins, the server exposes WCAG content as MCP tools your AI can call on demand. Want the exact wording of 2.4.7? It's there. Want the list of sufficient techniques for 1.3.1? It's there. Want to understand why 4.1.2 was reorganized in WCAG 2.2? It's there, in the Understanding doc, with the original phrasing intact.

This matters because accessibility is one of the few technical domains where the specification is both precise and contested. Citing the wrong criterion in a code review damages trust. Citing the wrong technique in a remediation plan wastes a sprint. Having the actual spec at the AI's fingertips means the conversation stops being about "what does Claude think WCAG says" and starts being about what WCAG actually says.

The audience is accessibility consultants, dev leads, and content writers who are tired of double-checking the AI's references against the W3C site every time a number gets quoted. If you're producing audit reports, training materials, or remediation tickets, this server pays for itself the first time it stops you from publishing a wrong rule number.

Made by joe-watkins. The <span class="whitespace-nowrap">a-gnt</span> Community catalogs it as essential reference plumbing for any AI workflow that touches WCAG. Pair with The Accessibility Auditor for full-page reviews, or WCAG Quick Audit for structured passes.