A11y MCP is a Model Context Protocol (MCP) server designed to provide Large Language Models (LLMs) with direct access to web accessibility testing capabilities. It serves developers, designers, and content creators who need to integrate automated accessibility checks into their workflows, enabling LLMs to programmatically assess web content for compliance with accessibility standards. The primary purpose is to bridge the gap between AI assistants and web accessibility evaluation, allowing for conversational and automated testing.
Web accessibility is a critical requirement for inclusive digital experiences, ensuring that websites and applications are usable by people with disabilities. However, manual accessibility testing is time-consuming and requires specialized knowledge. Many development teams struggle to consistently identify and fix accessibility issues, leading to non-compliant websites that exclude users. This server addresses the problem by automating accessibility testing through LLMs, making it easier to integrate checks into development pipelines and content creation processes.
One key feature is the ability to test web pages at any public URL for accessibility issues with customizable viewport dimensions. Users can specify width and height parameters to simulate different devices, such as mobile viewports like iPhone 12/13 (390x844 pixels) or default desktop viewports (1280x800 pixels). This allows for responsive design testing across various screen sizes, ensuring accessibility compliance on all devices.
Another major capability is testing raw HTML strings for accessibility issues. This feature enables developers to analyze code snippets, components, or dynamic content without deploying to a live URL. It supports the same customizable viewport and WCAG tag filtering as URL testing, making it ideal for testing individual UI components, email templates, or content management system outputs during development.
The server provides comprehensive WCAG compliance testing against various standards including WCAG 2.0, 2.1, and 2.2 at different conformance levels (A, AA, AAA). Users can specify which accessibility tags to test against, allowing for targeted compliance checks. Common tags include wcag2a, wcag2aa, wcag2aaa, wcag21a, wcag21aa, wcag22aa, and best-practice guidelines for comprehensive coverage.
Additional specialized tools include color contrast analysis to check if foreground and background color combinations meet WCAG requirements, ARIA attribute validation to test proper usage of accessibility attributes in HTML, and orientation lock detection to identify content that forces specific screen orientations. The server also offers rule exploration capabilities, allowing users to get information about available accessibility rules with optional filtering by tags.
admin
The server works by leveraging the Deque Axe-core accessibility engine through Puppeteer for browser automation. When an LLM requests an accessibility test, the server launches a headless browser, loads the specified URL or HTML content, runs the Axe-core analysis with configured parameters, and returns structured JSON results. This approach combines industry-standard accessibility testing libraries with modern browser automation to provide accurate, comprehensive testing results.
Users benefit from automated, programmatic accessibility testing that integrates directly with their LLM workflows. Developers can catch accessibility issues early in the development process, content creators can validate their work before publishing, and teams can maintain consistent accessibility standards across projects. The structured JSON output provides clear violation details, impact assessments, and remediation guidance, making it easier to prioritize and fix issues.
Concrete use cases include testing production websites for WCAG AA compliance during routine audits, validating HTML email templates for accessibility before sending campaigns, checking UI component libraries during design system development, analyzing competitor websites for accessibility benchmarking, testing single-page applications with dynamic content, and educating teams about accessibility requirements through interactive LLM conversations.
The target users are developers working with LLMs in Claude Desktop, Claude Code, VS Code, Cursor, and Windsurf environments. It integrates with MCP-compatible clients through standard configuration files and supports installation via npm with npx. The tech stack includes Node.js 18+, the Model Context Protocol SDK, Puppeteer for browser automation, and Axe-core for accessibility testing. The server is open-source under MIT license with over 6,000 downloads and available through various package managers.
A11y MCP provides LLMs with powerful web accessibility testing capabilities, enabling automated compliance checks, detailed issue identification, and comprehensive reporting through conversational interfaces and development workflows.
Developers and teams working with Large Language Models in MCP-compatible environments including Claude Desktop, Claude Code, VS Code, Cursor, and Windsurf. The server targets professionals who need to integrate automated web accessibility testing into their development workflows, content creation processes, or quality assurance pipelines. Users include frontend developers, accessibility specialists, UX designers, content creators, and QA engineers who require programmatic accessibility validation through conversational AI interfaces.