SitePulse MCP Server

SitePulse exposes a Model Context Protocol (MCP) server that lets AI coding assistants (Claude Code, Cursor, VS Code AI extensions) query your audit data, run audits, and retrieve recommendations directly from within their interfaces.

What is MCP?

The Model Context Protocol is an open standard for connecting AI assistants to external tools and data sources. By running SitePulse as an MCP server, your AI assistant can answer questions like:

• "What are the top SEO recommendations for acme-coffee.example?"
• "Run a URL audit on this page and summarize the results"
• "What is the current GEO score for my project?"

Available MCP tools

Setting up the MCP server

Step 1: Generate an API token

Go to Settings → API Keys → MCP Access Token and click Generate Token. This creates a long-lived token scoped to your organization's read/write operations.

Step 2: Configure your MCP client

Add SitePulse to your MCP client configuration. The server URL is:

https://app.sitepulse.app/mcp

For Claude Code (~/.claude.json):

{
  "mcpServers": {
    "sitepulse": {
      "type": "http",
      "url": "https://app.sitepulse.app/mcp",
      "headers": {
        "Authorization": "Bearer <your-mcp-token>"
      }
    }
  }
}

For Cursor or other clients, consult their MCP configuration documentation. The transport is HTTP (Server-Sent Events for streaming responses).

Authentication

MCP requests are authenticated via Bearer token in the Authorization header. The token is scoped to your organization and user account — it has the same permissions as your account role.

MCP tokens can be revoked at any time from Settings → API Keys → MCP Access Tokens.

Example usage in Claude Code

Once configured, you can ask Claude Code:

What does the latest audit for https://acme-coffee.example say about GEO?

Claude uses get_latest_audit to retrieve the audit, then summarizes the GEO score and top recommendations for you.

Or:

Run a URL audit on https://acme-coffee.example/blog/espresso-guide and tell me 
what JSON-LD schema is missing.

Claude uses run_audit (which may take 1–2 minutes), then analyzes the returned audit data for schema gaps.