Set up MCPs for GA4 and BigQuery so Claude can query your analytics data directly.

The End of the CSV Export

If you've ever exported a CSV from Google Analytics, cleaned it up in Sheets, and then pasted it into Claude just to ask a question about your traffic — this lesson is your liberation. We're going to connect Claude directly to GA4 and BigQuery so it can query your data in real time. No exports. No formatting. Just answers.

Prerequisites

Claude Desktop installed and running on your machine
A Google Cloud project with GA4 and BigQuery APIs enabled
A service account with read-only access (we'll create this together)

Step 1: Install the GA4 MCP

Open your Claude Desktop configuration file and add the GA4 and BigQuery MCP servers. On macOS, this file lives at ~/Library/Application Support/Claude/claude_desktop_config.json. Add the following to your mcpServers object:

claude_desktop_config.jsonjson

{
  "mcpServers": {
    "google-analytics": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-ga4"],
      "env": {
        "GA4_PROPERTY_ID": "properties/YOUR_PROPERTY_ID",
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
      }
    },
    "bigquery": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-bigquery"],
      "env": {
        "BQ_PROJECT_ID": "your-gcp-project-id",
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
      }
    }
  }
}

Step 2: Create a Read-Only Service Account

In Google Cloud Console, create a new service account with these roles:

Analytics Viewer (roles/analytics.viewer) — read-only access to GA4 reports
BigQuery Data Viewer (roles/bigquery.dataViewer) — read-only access to BigQuery tables
BigQuery Job User (roles/bigquery.jobUser) — permission to run queries

Watch Out

Never use your personal Google credentials for MCP connections. Always create a dedicated service account with the minimum permissions needed. This protects your account and makes it easy to revoke access if needed.

Restart and Verify

Quit Claude Desktop completely and reopen it. You should see the MCP tools icon (a small hammer) in the bottom-left of the chat input. Click it to verify that both google-analytics and bigquery servers are connected. If you see green checkmarks, you're ready to query.

Querying GA4

Prompt Example

claude

Claude queries the GA4 Data API directly, returning a formatted table with your real-time analytics data.

Using GA4, show me the top 10 landing pages by sessions this month. Include bounce rate and average session duration for each.

Querying BigQuery

Prompt Example

claude

Claude writes and executes a BigQuery SQL query, then presents the pathing analysis in a readable format.

Query BigQuery for the most common user journey paths — show me the top 5 sequences of page views that lead to a purchase event, with conversion rates for each path.

Try It Yourself

Connect Claude to your GA4 property and run your first live query.

Install the GA4 MCP using the configuration above, then ask Claude: 'What were our top 5 traffic sources last week, and how did each perform on conversion rate?'

Connecting Claude to Google Analytics & BigQuery