This server gives Claude five tools to audit web accessibility in HTML: check_alt_text scans images for missing or weak alt attributes, check_heading_hierarchy validates h1-h6 nesting and outputs a visual tree, check_color_contrast calculates WCAG contrast ratios for color pairs, validate_aria catches role misuse and missing properties, and check_form_accessibility ensures inputs have labels and fieldsets. You'd reach for this when reviewing markup for compliance issues or teaching an agent to spot common accessibility problems like unlabeled form fields, skipped heading levels, or aria-hidden on focusable elements. It runs over stdio and returns structured findings rather than browser automation.
Web accessibility analysis tools for AI agents, served over the Model Context Protocol (MCP).
Analyze images in HTML for missing, empty, or low-quality alt text.
alt attributesalt="" with role="presentation")<input type="image">, and <area> elementsValidate heading structure for proper nesting and content quality.
<h1> per pagerole="heading" with aria-levelCheck WCAG 2.x contrast ratio for foreground/background color pairs.
#fff, #ffffff), rgb(r,g,b), and named colorsValidate ARIA attributes on HTML elements.
aria-label on non-interactive elementsaria-hidden="true" on focusable elementsAudit form elements for accessibility compliance.
<fieldset> and <legend>autocomplete attributes for personal info fieldsnpm install
npm run build
Add to your claude_desktop_config.json:
{
"mcpServers": {
"a11y-tools": {
"command": "node",
"args": ["D:/products/mcp-servers/mcp-a11y-tools/dist/index.js"]
}
}
}
MIT