Accessible code at the speed of AI

Claude Code, Codex, Cursor, VS Code with GitHub Copilot, Gemini CLI, Windsurf, Zed, and Claude Desktop, along with any other host that supports Model Context Protocol. Developers keep working in the tools they already use.

The SDK finds accessibility issues; the MCP fixes them. The SDK is a testing tool that runs in your pipeline and reports what it detects, whether that’s a single component, a full page, or every deployment. AudioEye’s SDK MCP is a fixing tool that runs inside your AI coding agent, writing the correction and rescanning to confirm it worked. They solve different halves of the same problem.

Yes. Approval is what keeps you in control of your own codebase. The agent has already found the issue, written the correction, and can confirm it worked, so your developer is reading a finished change rather than researching one. They approve it and ship it through their normal process, which means the fix moves at the speed of any other code review.

No. The MCP uses automation to detect and fix issues in your code, and automation cannot catch every accessibility issue. Expert audits test with assistive technology and real users, which is how the high-impact issues get found. Most teams need both.

Not reliably. Most coding agents learned from a web that mostly isn't accessible, so unprompted, they default to div buttons, unlabeled inputs, and missing alt text. They also can't see the rendered page, so anything that only appears at runtime, like focus order, contrast in context, or screen reader behavior, goes unnoticed.

Yes. AudioEye's SDK MCP connects your coding agent to real accessibility testing, so it can scan a page, pull back specific WCAG issues with the elements and success criteria involved, and return recommended fixes without leaving the IDE. You review and apply the changes.

An important note: no automated tool certifies compliance on its own. The SDK MCP handles the automated layer at the point of authoring. Expert testing handles what requires human judgment.

You'll need an AudioEye account with Testing SDK access, Node.js 22 or newer, and a web app you can run locally. Then:

1. Run npx -y @audioeye/testing-sdk-mcp@latest setup once and restart your coding agent. Setup registers the server with Claude Code, Codex, Cursor, VS Code, Gemini CLI, Windsurf, and Zed in one pass. Claude Code users can install it as a plugin instead, with no terminal.

2. Ask the agent to sign in to AudioEye. It returns a pairing link; approve it in the browser once.

3. Start your dev server and run /mcp__audioeye__scan-and-fix http://localhost:3000.

For CI, Claude Desktop, and per-host details, see our developer docs(opens in a new tab).

Model Context Protocol is an open standard that connects an AI agent to an outside source of data or expertise. An MCP server is what provides that source. AudioEye's SDK MCP, for example, is an MCP server that gives your coding agent access to AudioEye's accessibility data set.