What is the AccessiSight MCP server?

It's a Model Context Protocol server (at /api/mcp) that exposes AccessiSight's accessibility scanning engine as tools an AI coding agent can call directly (scan_website, get_scan_results, generate_ai_fix, and list_wcag_coverage), so agents like Claude Code or Cursor can check and fix accessibility issues during a normal coding session.

How do I connect the AccessiSight MCP server to Claude Code or Cursor?

Add an MCP server entry pointing to https://accessisight.com/api/mcp in your agent's MCP config (e.g. Claude Code's mcpServers config or Cursor's MCP settings). Once connected, the agent can call scan_website, get_scan_results, generate_ai_fix, and list_wcag_coverage directly.

Do I need a paid plan to use the MCP server?

get_scan_results, generate_ai_fix, and list_wcag_coverage are open to any caller. scan_website (starting a new scan) requires a Pro license key passed as a licenseKey argument.

Developer ToolsNEW

MCP Server

Connect Claude Code, Claude Desktop, Cursor or any MCP-compatible client directly to AccessiSight's real scanner and Smart Fix, with no local install and no separate API client to write.

Claude Code / Claude Desktop

Add this to your .mcp.json (project-level) or Claude Desktop's MCP server config:

{
  "mcpServers": {
    "accessisight": {
      "url": "https://accessisight.com/api/mcp"
    }
  }
}

Cursor

Add this to your .cursor/mcp.json:

{
  "mcpServers": {
    "accessisight": {
      "url": "https://accessisight.com/api/mcp"
    }
  }
}

Available Tools

scan_website

Starts a real accessibility scan of any URL with the same in-house scanner engine that powers the web app (over 450 rules in 31 groups, including real keyboard testing; no axe-core and no third-party rules engine). Returns a scanId immediately; the scan runs in the background.

Requires: A Pro license key (see Pricing), passed as a licenseKey argument.

get_scan_results

Polls a scan's status by scanId. Returns the real summary (severity counts, score) and violation list once complete, including a captured HTML snippet per violation where available.

Requires: Nothing. Open to any caller.

generate_ai_fix

Generates a code fix for a specific violation (a description and the offending HTML are required). It tries Claude, falls back to Gemini, then to a deterministic template if neither AI provider is configured on the server. The response says which engine produced the fix.

Requires: Nothing. Open to any caller.

list_wcag_coverage

Returns the breakdown of what this scanner can and cannot detect automatically across all WCAG 2.2 AA success criteria, including the reason for each of the ~13 criteria that need human judgment. Useful for an agent to understand this tool's limits before relying on it.

Requires: Nothing. Open to any caller.

Example agent workflow

Once connected, you can ask your AI coding assistant things like:

"Scan my staging site at staging.example.com and tell me the top 3 critical issues."
"Generate a fix for the color contrast violation on the login button."
"What WCAG criteria can't be automatically checked, and why?"

Frequently Asked Questions

Ready to Find Real Issues?

Skip the setup: run a free accessibility scan on your own site right now, with nothing to install.

Run a Free Scan