Connect SiteCMD to Cursor so your AI assistant can read scan results, list issues, and generate fix prompts without leaving the editor.

Configuration

The easiest way is to let the desktop app do it: open SiteCMD, go to Integrations, and connect Cursor. The app writes the config and runs the MCP server bundled inside the app via your local Node, so there’s nothing to install and the script path is resolved for your OS.

To wire it up by hand, add the bundled server script to ~/.cursor/mcp.json (create the file if it doesn’t exist):

{
  "mcpServers": {
    "sitecmd": {
      "command": "node",
      "args": [
        "--disable-warning=ExperimentalWarning",
        "/Applications/SiteCMD.app/Contents/Resources/sitecmd-mcp/sitecmd-mcp.mjs"
      ]
    }
  }
}

That path is the macOS location. On Windows and Linux the script lives in the resources/sitecmd-mcp/ folder next to the installed SiteCMD binary; letting the app write the config is the reliable way to get it right. Restart Cursor after saving.

What you get

Once connected, Cursor’s AI can call these tools against your SiteCMD data:

get_projects - list all projects tracked in SiteCMD
get_scan_score - fetch the latest score and category breakdown
get_issues - list failing issues ranked by severity and impact
get_fix_prompts - generate fix prompts for selected issues
get_scan_history - retrieve score history over time
get_dismissed_issues - review dismissed or not-applicable issues
compare_scans - compare two scans for fixed, new, and still-open issues
request_scan - get instructions for running a scan and comparing results