How to Connect Google Search Console to Claude with MCP

If you have ever exported Search Console data to a spreadsheet just to answer "which queries lost clicks last week?", there is now a much shorter path: give your AI assistant a direct, read-only connection to your Search Console data and ask in plain language.

This guide shows how to do that with the Model Context Protocol (MCP) and a hosted connection from KumoMCP — no local server, no API keys, no code.

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants like Claude, Cursor and ChatGPT securely call external tools and data sources. An MCP server exposes tools (for example, "run a search analytics query"); the assistant decides when to call them while answering your questions.

You can run an MCP server yourself, but that means installing software, managing OAuth credentials and keeping it running. A hosted MCP service does that part for you: you sign in, authorize Google once, and get a personal MCP URL to paste into your assistant.

Step 1: Sign in and authorize Google

Create a KumoMCP account and connect your Google account with read-only Search Console permission. KumoMCP never asks for write access, and OAuth tokens are encrypted with AES-256-GCM before they touch the database.

Step 2: Pick your property

Choose which Search Console property your assistant should see. The free plan includes one property, one MCP endpoint and every data tool.

Step 3: Paste your MCP URL

Copy your personal MCP URL from the dashboard and add it to your client:

• Claude Code: claude mcp add kumomcp <your-url>
• Claude Desktop / Claude.ai: add a custom connector and paste the URL
• Cursor: add an entry under MCP settings with the URL
• ChatGPT: add the URL as a connector in settings

That's it. There is no step 4.

What you can ask

Once connected, the assistant picks the right tool for each question:

• Which queries lost the most clicks in the last 28 days?
• Compare mobile and desktop CTR for my top 20 pages this month.
• Is /pricing indexed, and when was it last crawled?
• Do any of my sitemaps have errors right now?

Every tool is read-only, so the assistant can analyze your data but can never change anything in your Google account.

Why hosted beats local here

A local MCP server works, but a hosted endpoint has three practical wins: it works in web clients like Claude.ai and ChatGPT that cannot launch local processes; the same URL works on every device; and there is nothing to install, update or babysit. If your endpoint URL ever leaks, rotate it from the dashboard with one click.

Ready to try it? Start free — one property, every tool, no credit card.