accessibilitymcp
Enables LLMs to perform web accessibility testing via the MCP protocol by invoking Axe-core.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: accessibilitymcp. It is used for: Enables LLMs to perform web accessibility testing via the MCP protocol by invoking Axe-core. Full Skill content: https://321skill.com/skills/accessibilitymcp/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This Skill addresses the challenge of integrating web accessibility testing into AI development workflows. In practice, developers frequently need to check if pages comply with WCAG standards, but traditional tools (like the Axe DevTools browser extension) require manual operation and cannot be triggered directly within a conversation. Through the MCP protocol, an LLM can directly invoke the Axe-core engine and Puppeteer to automatically analyze any public URL or HTML snippet, returning a detailed violation report that includes impact level, remediation guidance, and WCAG guideline references.
Usage is straightforward: simply configure the server in an MCP-compatible client (such as Claude Desktop, VS Code, or Cursor), after which you can initiate tests using natural language instructions. For example, you can say, "Test this website https://example.com for WCAG 2.1 AA compliance." It will automatically launch a headless browser to load the page, run the Axe-core analysis, and return a list of violations, each accompanied by a description, suggested fix, and the corresponding WCAG guideline. You can also directly test an HTML string without needing to deploy the page.
It is ideal for front-end development teams, QA engineers, and product managers who need to ensure web accessibility compliance—particularly teams already using AI coding assistants for development and wishing to seamlessly integrate accessibility testing into their daily conversations. For large projects, you can ask the AI to check the current page for accessibility issues while developing new features, preventing rework later.
We recommend using this tool for unified accessibility checks during new feature development or page refactoring, and incorporating the results into the code review process. Note that it primarily tests individual pages or HTML snippets; it cannot handle pages requiring login authentication or complex single-page application (SPA) routing (e.g., dynamic routes within an SPA). Test results are also affected by page load state. The initial run will automatically download Chromium, which may take some time.
Key Features
Unlike the traditional Axe DevTools browser extension, it leverages the MCP protocol to allow LLMs to invoke tests directly, enabling automated testing within a conversation without manually opening developer tools. Compared to Lighthouse, it offers more granular rule exploration (such as listing all available rules and customizing test tags), specialized features like color contrast analysis and ARIA validation, and supports direct testing of HTML snippets.
Limitations
Requires Node.js 18+. Relies on Puppeteer to automatically download Chromium (slow initial startup). Only supports public URLs or HTML strings; cannot test pages requiring login authentication or SPA internal routing.
FAQ
Which WCAG versions are supported?
Supports WCAG 2.0, 2.1, and 2.2. You can specify the version via instruction, e.g., 'Check for WCAG 2.1 AA compliance.'
How do I test an HTML file under local development?
You can directly paste the HTML code into the conversation and use the 'Test this HTML' instruction; no local server startup is required.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/accessibilitymcp/raw/index.md to read the original Skill definition (Markdown format) for accessibilitymcp, and install it according to the instructions.
Raw Markdown URL for AI: /skills/accessibilitymcp/raw/index.md