Google Analytics MCP Server

MCP Google Analytics Server

A Model Context Protocol (MCP) server for Google Analytics 4, providing comprehensive integration with both the Google Analytics Data API (for reading reports) and Measurement Protocol v2 (for sending events).

The GA4 MCP that reads AND writes. Most GA4 MCP servers (including Google's official one) are read-only. This one gives your AI agent the full loop: run reports and funnels, audit your setup (custom dimensions, key events, compatibility checks), send ecommerce and conversion events server-side, and verify them in the realtime report — 26 tools in one npx command.

Built for agencies too: every read tool accepts an optional propertyId, so one conversation can query all your clients' properties — no reconfiguration between clients. See Multi-Property Mode.

⚡ One-Click Install

Click Install in Cursor above, approve, then replace the placeholder values in ~/.cursor/mcp.json with your real credentials (see Configuration below).

⚡ Token Optimization - READ THIS FIRST!

IMPORTANT: Google Analytics reports can return large datasets that consume significant tokens. This server is designed with token optimization in mind:

• All read tools default to 10 results - Adjust the limit parameter as needed
• Use specific date ranges - Avoid querying years of data at once
• Select only needed dimensions/metrics - Don't request everything
• Check TOKEN_OPTIMIZATION.md for detailed best practices

See the dedicated Token Optimization Guide for strategies to minimize token usage.

🚀 Quick Start

See QUICKSTART.md for a 5-minute setup guide, or follow the installation steps below.

📦 Installation

Option 1: Install globally via npm

npm install -g mcp-google-analytics

Option 2: Use with npx (no installation needed)

npx mcp-google-analytics

🔧 Configuration

This server requires different credentials for reading data vs sending events:

For Reading Data (Google Analytics Data API)

You need a Service Account with access to your GA4 property:

1. Go to Google Cloud Console
2. Create or select a project
3. Enable the Google Analytics Data API
4. Create a Service Account:
   • Go to "IAM & Admin" > "Service Accounts"
   • Click "Create Service Account"
   • Give it a name (e.g., "GA4 MCP Reader")
   • Grant the "Viewer" role
   • Create a JSON key and download it
5. Add the service account email to your GA4 property:
   • Go to GA4 Admin > Property Access Management
   • Add the service account email with "Viewer" role
6. Get your Property ID:
   • Go to GA4 Admin > Property Settings
   • Copy the Property ID (numeric, e.g., "123456789")

For Sending Events (Measurement Protocol)

You need a Measurement ID and API Secret:

1. Go to GA4 Admin > Data Streams
2. Select your data stream (web, iOS, or Android)
3. Copy the Measurement ID (format: G-XXXXXXXXXX)
4. Click "Measurement Protocol API secrets"
5. Click "Create" to generate a new API secret
6. Copy the secret value

Environment Variables

Set these environment variables:

# For Data API (reading)
export GA_SERVICE_ACCOUNT_JSON=/path/to/service-account.json
# Or provide JSON directly:
# export GA_SERVICE_ACCOUNT_JSON='{"type":"service_account","project_id":"..."}'

# Optional: default property. If omitted, pass "propertyId" per tool call
# (multi-property mode — see below)
export GA_PROPERTY_ID=123456789

# For Measurement Protocol (writing)
export GA_MEASUREMENT_ID=G-XXXXXXXXXX
export GA_API_SECRET=your-api-secret-here

🏢 Multi-Property Mode (Agencies)

Every read tool accepts an optional propertyId argument that overrides the configured GA_PROPERTY_ID — so a single conversation can query any property the service account can access, with no reconfiguration between clients.

1. Grant your service account "Viewer" access on each client's GA4 property (or at account level).
2. Set only GA_SERVICE_ACCOUNT_JSON (GA_PROPERTY_ID becomes optional — if set, it acts as the default).
3. Discover properties, then query any of them:

Show me all my accounts and properties          → ga_get_account_summaries
Compare last week's active users between the    → ga_run_report with propertyId "111111"
Acme property and the Globex property             and again with propertyId "222222"

propertyId accepts both 123456789 and properties/123456789. Event sending (Measurement Protocol) remains tied to the configured GA_MEASUREMENT_ID/GA_API_SECRET, since each data stream has its own secret.

🔌 Integration with Claude Desktop

Add to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "google-analytics": {
      "command": "npx",
      "args": ["-y", "mcp-google-analytics"],
      "env": {
        "GA_SERVICE_ACCOUNT_JSON": "/path/to/service-account.json",
        "GA_PROPERTY_ID": "123456789",
        "GA_MEASUREMENT_ID": "G-XXXXXXXXXX",
        "GA_API_SECRET": "your-api-secret"
      }
    }
  }
}

Or if installed globally:

{
  "mcpServers": {
    "google-analytics": {
      "command": "mcp-google-analytics",
      "env": {
        "GA_SERVICE_ACCOUNT_JSON": "/path/to/service-account.json",
        "GA_PROPERTY_ID": "123456789",
        "GA_MEASUREMENT_ID": "G-XXXXXXXXXX",
        "GA_API_SECRET": "your-api-secret"
      }
    }
  }
}

Restart Claude Desktop after updating the configuration.

🎯 Integration with Cursor

Add to your Cursor MCP settings file:

macOS/Linux: ~/.cursor/mcp.json Windows: %USERPROFILE%\.cursor\mcp.json

{
  "mcpServers": {
    "google-analytics": {
      "command": "npx",
      "args": ["-y", "mcp-google-analytics"],
      "env": {
        "GA_SERVICE_ACCOUNT_JSON": "/path/to/service-account.json",
        "GA_PROPERTY_ID": "123456789",
        "GA_MEASUREMENT_ID": "G-XXXXXXXXXX",
        "GA_API_SECRET": "your-api-secret"
      }
    }
  }
}

Restart Cursor after updating the configuration.

🛠️ Available Tools

At a Glance

Reading data (Google Analytics Data API) — all read tools default to 10 rows to save tokens:

Sending events (Measurement Protocol):

Google Analytics Data API (Reading Data)

ga_run_report

Run custom reports with dimensions and metrics.

Common Dimensions: date, city, country, deviceCategory, browser, pagePath, eventName, sessionSource, sessionMedium, sessionCampaignName

Common Metrics: activeUsers, sessions, screenPageViews, conversions, totalRevenue, engagementRate, averageSessionDuration

Example:

{
  "dateRanges": [{"startDate": "7daysAgo", "endDate": "today"}],
  "dimensions": [{"name": "city"}],
  "metrics": [{"name": "activeUsers"}],
  "limit": 10
}

ga_run_realtime_report

Get real-time data (last 30 minutes).

Example:

{
  "metrics": [{"name": "activeUsers"}],
  "dimensions": [{"name": "country"}],
  "limit": 10
}

ga_get_metadata

Get all available dimensions and metrics for your property.

Warning: Returns 500+ items. Use sparingly.

ga_list_accounts

List all GA accounts accessible to the service account.

ga_list_properties

List GA4 properties, optionally filtered by account ID.

ga_get_property

Get details about the configured property.

ga_list_data_streams

List data streams for the configured property.

ga_run_pivot_report

Run pivot table reports with row/column dimensions.

Example:

{
  "dateRanges": [{"startDate": "7daysAgo", "endDate": "today"}],
  "dimensions": [{"name": "country"}, {"name": "deviceCategory"}],
  "metrics": [{"name": "activeUsers"}],
  "pivots": [{"fieldNames": ["deviceCategory"], "limit": 5}]
}

ga_run_funnel_report

Run funnel analysis to track user progression.

Note: Funnel reporting uses the Data API v1alpha channel (the only channel where Google exposes it). Each step matches an event: set eventName per step, or omit it to use the step's name as the event name. For advanced matching, pass a full filterExpression.

Example:

{
  "dateRanges": [{"startDate": "7daysAgo", "endDate": "today"}],
  "funnelSteps": [
    {"name": "page_view"},
    {"name": "add_to_cart"},
    {"name": "begin_checkout"},
    {"name": "Purchase", "eventName": "purchase"}
  ]
}

ga_batch_run_reports

Run multiple reports in a single request.

Warning: Can return large datasets. Limit to 2-5 reports per batch.

ga_get_account_summaries

Get all accessible accounts with their properties in a single compact call. The fastest way to discover account and property IDs.

ga_list_custom_dimensions / ga_list_custom_metrics

List the custom dimensions and metrics defined for the property, including their API names (e.g., customEvent:plan_type) so you can use them in reports.

ga_list_key_events

List the key events (conversions) configured for the property — useful before building conversion reports or deciding which events to send.

ga_list_google_ads_links

List Google Ads accounts linked to the property.

ga_check_compatibility

Check whether a dimension/metric combination is valid before running a report, avoiding wasted requests and token-heavy error loops.

Example:

{
  "dimensions": [{"name": "city"}],
  "metrics": [{"name": "activeUsers"}],
  "compatibilityFilter": "COMPATIBLE"
}

Measurement Protocol (Sending Events)

Good to know:

• Events take a few minutes to appear in standard reports, but show up almost immediately in the realtime report (ga_run_realtime_report).
• Use ga_validate_event to test new events without recording them.
• If you omit client_id, one is auto-generated per call. To have GA group several events (e.g., a cart-to-purchase flow) into the same session and user, pass the same client_id to every call.

ga_send_event

Send custom events to GA4.

Example:

{
  "events": [{
    "name": "button_click",
    "params": {
      "button_id": "cta_signup",
      "page": "/landing"
    }
  }],
  "user_id": "user123"
}

ga_validate_event

Validate events before sending (uses debug endpoint).

ga_send_pageview

Send page view events.

Example:

{
  "page_location": "https://example.com/products",
  "page_title": "Products",
  "user_id": "user123"
}

ga_send_purchase

Send ecommerce purchase events.

Example:

{
  "transaction_id": "T12345",
  "value": 99.99,
  "currency": "USD",
  "items": [{
    "item_id": "SKU123",
    "item_name": "Product Name",
    "price": 99.99,
    "quantity": 1
  }]
}

ga_send_login

Send login events.

ga_send_signup

Send user registration events.

ga_send_view_item

Send product/item detail view events. Completes the standard ecommerce funnel: view_item → add_to_cart → begin_checkout → purchase.

ga_send_add_to_cart

Send add-to-cart events.

ga_send_begin_checkout

Send checkout initiation events.

ga_send_refund

Send full or partial refund events. Use the same transaction_id as the original purchase; omit items for a full refund, include them for a partial one.

Example (partial refund):

{
  "transaction_id": "T12345",
  "currency": "USD",
  "value": 49.99,
  "items": [{"item_id": "SKU123", "quantity": 1}]
}

📖 Usage Examples

See EXAMPLES.md for practical usage examples in Spanish.

Example: Get users by country (last 7 days)

Show me active users by country for the last 7 days

Claude will use ga_run_report:

{
  "dateRanges": [{"startDate": "7daysAgo", "endDate": "today"}],
  "dimensions": [{"name": "country"}],
  "metrics": [{"name": "activeUsers"}],
  "limit": 10,
  "orderBys": [{"metric": {"metricName": "activeUsers"}, "desc": true}]
}

Example: Track a purchase

Send a purchase event for order #12345, $99.99 USD

Claude will use ga_send_purchase:

{
  "transaction_id": "12345",
  "value": 99.99,
  "currency": "USD",
  "items": [{
    "item_id": "product_1",
    "item_name": "Example Product",
    "price": 99.99,
    "quantity": 1
  }]
}

Example: Validate an event before sending it

Recommended before wiring up any new event: the debug endpoint checks the payload without recording anything.

Validate this tutorial_complete event before we send it for real

Claude will use ga_validate_event:

{
  "client_id": "test.123",
  "events": [{
    "name": "tutorial_complete",
    "params": {"tutorial_id": "onboarding", "duration_seconds": 120}
  }]
}

The response lists validation messages; an empty list means the event is well-formed.

Example: Server-side conversion tracking

Track signups or logins that happen in your backend, where no JavaScript tag runs:

A user just registered with Google OAuth, record the signup in Analytics

Claude will use ga_send_signup:

{
  "user_id": "user_789",
  "method": "Google"
}

Example: Full ecommerce funnel from an agent

Send the same client_id on each call so GA groups the events into one session:

Track this user's journey: they added a $49 course to the cart, started checkout, and completed the purchase

Claude will chain ga_send_add_to_cart → ga_send_begin_checkout → ga_send_purchase, reusing the client ID:

{
  "client_id": "555.1717000000",
  "currency": "USD",
  "value": 49,
  "items": [{"item_id": "course_101", "item_name": "Intro Course", "price": 49, "quantity": 1}]
}

Example: Custom events from automations

Measure things GA never sees natively, like AI agent activity or scheduled jobs:

Log that the weekly report generator ran successfully

Claude will use ga_send_event:

{
  "events": [{
    "name": "automation_run",
    "params": {"job": "weekly_report", "status": "success", "duration_ms": 5400}
  }]
}

Example: Send and verify in one conversation

Combine both APIs to confirm your tracking works end to end:

Send a test event and confirm Analytics received it

Claude will call ga_send_event, then check with ga_run_realtime_report:

{
  "dimensions": [{"name": "eventName"}],
  "metrics": [{"name": "eventCount"}],
  "limit": 10
}

Measurement Protocol events appear in the realtime report within seconds, while standard reports can take a few minutes.

🔍 Debugging

Enable debug logging by setting:

export DEBUG=mcp-google-analytics:*

For Measurement Protocol, use ga_validate_event to check events before sending them live.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

MIT License - see LICENSE file for details.

🔗 Links

• GitHub Repository
• npm Package
• Smithery
• MCP Documentation
• GA4 Data API Documentation
• Measurement Protocol Documentation

🆘 Support

For issues and questions:

• GitHub Issues
• MCP Community Discord

📝 Changelog

See CHANGELOG.md for version history.

---