Integrations · MCP server

Let your AI agent run compliance scans for you

We ship a Model Context Protocol (MCP) server that exposes the entire scan API as tools an AI assistant can call. Add it once to Claude Desktop, Cursor, Zed, or any MCP-aware client, and your agent can list domains, queue scans, and read results without leaving the conversation.

Tools exposed

list_domains — every domain on your account with verification status
add_domain — register a new hostname and receive the meta tag to verify it
verify_domain — re-check the verification token after deploying the meta tag
run_scan — queue any subset of the four pillars against a verified domain
get_scan — fetch the latest status and per-pillar scores
list_scans — recent scan history

Setup

Generate an API key from your settings page.
Add this entry to your MCP client config (Claude Desktop:claude_desktop_config.json):

{
  "mcpServers": {
    "auditly": {
      "command": "npx",
      "args": ["-y", "@auditly-fyi/mcp"],
      "env": {
        "AUDITLY_API_KEY": "sgk_...",
        "AUDITLY_BASE_URL": "https://your-deploy.example.com"
      }
    }
  }
}

Example conversation

You

Run a fresh scan on all my verified domains and tell me which one has the lowest accessibility score.

Assistant

Calls list_domains → gets 4 verified domains → calls run_scan on each with all 4 pillars → polls get_scan until each finishes → reports back the rankings, with the lowest-scoring site and the top 3 violations to fix.

The MCP package source lives in apps/mcp/. It's a thin translation layer over the same REST API anyone else can use.