Connect Google Search Console to Claude

What you'll be able to do: ask Claude things like "Which pages lost traffic this month?" or "Which keywords am I close to ranking on page 1 for?" and get answers from your own Search Console data, in plain English.

Time: about 5 minutes · Cost: free · Level: beginner (copy and paste)

Before you start

• Node.js v18 or higher (check by running node -v in Terminal)
• A Google account with Search Console access
• A Claude account (free or paid)

Phase 1: Download Claude Desktop

1. Go to claude.ai.
2. Click your profile icon (top right) → Download Claude Desktop.
3. Install it and log in.

Phase 2: Set up Google Cloud credentials

This gives Claude read-only access to your Search Console data through Google's official login.

4. Go to console.cloud.google.com.

5. Click "Select a project" → "New Project". Name it GSC-MCP and click Create.

6. In the left sidebar, go to APIs & Services → Library. Search for "Google Search Console API", open it and click Enable.

7. Go to APIs & Services → OAuth consent screen. Select External → Create. Fill in:

   • App name: GSC MCP
   • User support email: your Gmail
   • Developer contact email: your Gmail

   Click Save and Continue through the remaining steps (defaults are fine).

8. Go to APIs & Services → Credentials. Click "+ Create Credentials" → OAuth client ID.

   • Application type: Desktop app
   • Name: GSC MCP Client
   • Click Create, then Download JSON in the popup.
   • Save the file somewhere easy to find, e.g. ~/Documents/gsc-credentials.json

9. Back on the OAuth consent screen, scroll down → "+ Add Users", add your own Google email and click Save.

⚠️ This step is required because the app is in test mode. You must add yourself as a test user.

Phase 3: Configure Claude Desktop

10. Open Claude Desktop.
11. Go to Settings → Developer → Edit Config. This opens claude_desktop_config.json.
12. Add this block inside "mcpServers": { }, just before the closing }:

"gsc": {
  "command": "npx",
  "args": ["-y", "suganthan-gsc-mcp"],
  "env": {
    "GSC_AUTH_MODE": "oauth",
    "GSC_OAUTH_SECRETS_FILE": "/Users/YOUR-USERNAME/Documents/gsc-credentials.json",
    "GSC_SITE_URL": "sc-domain:yoursite.com"
  }
}

Replace two things:

• /Users/YOUR-USERNAME/Documents/gsc-credentials.json → the real path to the JSON file from step 8
• yoursite.com → your domain as it appears in Search Console

13. Save the file.
14. Fully quit Claude Desktop (Mac: Cmd + Q. Windows: right-click the taskbar icon → Quit).
15. Reopen Claude Desktop.

Phase 4: Connect and sign in

16. Open a new chat.
17. Type: List my Search Console properties
18. A browser window opens. Sign in with Google and allow access.

✅ This only happens once. Your login is saved after the first time.

You're connected. Try asking:

• "What are my top 10 keywords by impressions this month?"
• "Which pages have high impressions but low click-through rate?"
• "Which keywords am I ranking between position 8 and 20?" (quick wins)
• "Are any of my pages competing for the same keyword?"
• "Check the indexing status of this URL: yoursite.com/page"
• "Which pages have lost the most traffic in the last 3 months?"

Bonus prompt: find quick click-through wins

You have access to my Google Search Console data. Run a "seen but not clicked" audit to find quick CTR wins.

STEP 1 — Pull the data
Pull queries and their landing pages from the last 90 days.

STEP 2 — Filter
Keep only rows that are getting impressions but few or no clicks — i.e. I'm showing in search but people aren't clicking through.

STEP 3 — Pick the top 10
Sort by impressions, most-seen first. Take the top 10 opportunities.

STEP 4 — Rewrite for each
For every one, write a new title tag and meta description built to earn the click:
- Title tag: under 60 characters, work the main query in naturally, lead with a clear benefit or hook, no clickbait.
- Meta description: under 155 characters, expand on the title, include the query once, end with a reason to click.

OUTPUT
Give me ONE table with these exact columns:
| Page URL | Top Query | Impressions | Clicks | Suggested Title Tag | Suggested Meta Description |

Use the real landing page URL from GSC for each row (not the query).
After the table, add one line per row briefly explaining what was weak about the likely current title and why the rewrite should do better.