Google Analytics MCP Server

mcp-name: io.github.surendranb/google-analytics-mcp

Connect Google Analytics 4 data to AI agents, agentic workflows, and MCP clients. Give agents analysis-ready access to website traffic, user behavior, and performance data with schema discovery, server-side aggregation, and safe defaults that reduce data wrangling.

Built for: AI agents, analyst copilots, and MCP runtimes across Claude, ChatGPT, Cursor, Windsurf, and custom hosts.

I also built a Google Search Console MCP that enables you to mix & match the data from both the sources

---

Why Agents Use This Server

• Analysis-ready outputs with server-side aggregation, so agents spend more time answering questions and less time wrangling rows
• Live schema discovery for each GA4 property, including category-based exploration for dimensions and metrics
• Context-safe defaults that estimate large datasets before they blow up a conversation or workflow
• Portable MCP surface that works across agent runtimes, IDE copilots, and custom automation

---

Prerequisites

Check your Python setup:

# Check Python version (need 3.10+)
python --version
python3 --version

# Check pip
pip --version
pip3 --version

Required:

• Python 3.10 or higher
• Google Analytics 4 property with data
• Service account with Google Analytics Data API access and GA4 property access

---

Step 1: Setup Google Analytics Credentials

Create Service Account in Google Cloud Console

1. Go to Google Cloud Console
2. Create or select a project:
   • New project: Click "New Project" → Enter project name → Create
   • Existing project: Select from dropdown
3. Enable the Analytics APIs:
   • Go to "APIs & Services" → "Library"
   • Search for "Google Analytics Data API" → Click "Enable"
4. Create Service Account:
   • Go to "APIs & Services" → "Credentials"
   • Click "Create Credentials" → "Service Account"
   • Enter name (e.g., "ga4-mcp-server")
   • Click "Create and Continue"
   • Skip role assignment → Click "Done"
5. Download JSON Key:
   • Click your service account
   • Go to "Keys" tab → "Add Key" → "Create New Key"
   • Select "JSON" → Click "Create"
   • Save the JSON file - you'll need its path

Add Service Account to GA4

1. Get service account email:
   • Open the JSON file
   • Find the client_email field
   • Copy the email (format: ga4-mcp-server@your-project.iam.gserviceaccount.com)
2. Add to GA4 property:
   • Go to Google Analytics
   • Select your GA4 property
   • Click "Admin" (gear icon at bottom left)
   • Under "Property" → Click "Property access management"
   • Click "+" → "Add users"
   • Paste the service account email
   • Select "Viewer" role
   • Uncheck "Notify new users by email"
   • Click "Add"

Find Your GA4 Property ID

1. In Google Analytics, select your property
2. Click "Admin" (gear icon)
3. Under "Property" → Click "Property details"
4. Copy the Property ID (numeric, e.g., 123456789)
   • Note: This is different from the "Measurement ID" (starts with G-)

Test Your Setup (Optional)

Verify your credentials:

pip install google-analytics-data

Create a test script (test_ga4.py):

import os
from google.analytics.data_v1beta import BetaAnalyticsDataClient

# Set credentials path
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "/path/to/your/service-account-key.json"

# Test connection
client = BetaAnalyticsDataClient()
print("✅ GA4 credentials working!")

Run the test:

python test_ga4.py

If you see "✅ GA4 credentials working!" you're ready to proceed.

---

Step 2: Install the MCP Server

There are two supported ways to launch the server:

• ga4-mcp-server when the installed console script is available on your PATH
• python -m ga4_mcp when you want to use a specific interpreter or virtual environment

Method A: Install from PyPI (Recommended)

python3 -m pip install google-analytics-mcp

If your machine uses python instead of python3, run:

python -m pip install google-analytics-mcp

Option 1: Use the console script

Use this when ga4-mcp-server is available on your PATH:

{
  "mcpServers": {
    "ga4-analytics": {
      "command": "ga4-mcp-server",
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/service-account-key.json",
        "GA4_PROPERTY_ID": "123456789"
      }
    }
  }
}

Option 2: Use an explicit Python interpreter

Use this when you want to pin the exact Python runtime or when the console script is not on your PATH.

If python3 --version worked:

{
  "mcpServers": {
    "ga4-analytics": {
      "command": "python3",
      "args": ["-m", "ga4_mcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/service-account-key.json",
        "GA4_PROPERTY_ID": "123456789"
      }
    }
  }
}

If python --version worked:

{
  "mcpServers": {
    "ga4-analytics": {
      "command": "python",
      "args": ["-m", "ga4_mcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/service-account-key.json",
        "GA4_PROPERTY_ID": "123456789"
      }
    }
  }
}

Method B: Install from a local clone

git clone https://github.com/surendranb/google-analytics-mcp.git
cd google-analytics-mcp
python3 -m venv .venv
source .venv/bin/activate
python -m pip install .

If you plan to modify the package locally, use python -m pip install -e . instead.

MCP Configuration:

{
  "mcpServers": {
    "ga4-analytics": {
      "command": "/full/path/to/google-analytics-mcp/.venv/bin/python",
      "args": ["-m", "ga4_mcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/service-account-key.json",
        "GA4_PROPERTY_ID": "123456789"
      }
    }
  }
}

---

Step 3: Update Configuration

Replace these placeholders in your MCP configuration:

• /path/to/your/service-account-key.json with the absolute path to your JSON key
• 123456789 with your numeric GA4 Property ID
• /full/path/to/google-analytics-mcp/.venv/bin/python with your virtual environment's Python path (Method B only)

---

Usage

Once configured, ask your MCP client questions like:

Discovery & Exploration

• What GA4 dimension categories are available?
• Show me all ecommerce metrics
• What dimensions can I use for geographic analysis?

Traffic Analysis

• What's my website traffic for the past week?
• Show me user metrics by city for last month
• Compare bounce rates between different date ranges

Multi-Dimensional Analysis

• Show me revenue by country and device category for last 30 days
• Analyze sessions and conversions by campaign and source/medium
• Compare user engagement across different page paths and traffic sources

E-commerce Analysis

• What are my top-performing products by revenue?
• Show me conversion rates by traffic source and device type
• Analyze purchase behavior by user demographics

---

Quick Start Examples

Try these example queries to see the MCP's analytical capabilities:

1. Geographic Distribution

Show me a map of visitors by city for the last 30 days, with a breakdown of new vs returning users

This demonstrates:

• Geographic analysis
• User segmentation
• Time-based filtering
• Data visualization

2. User Behavior Analysis

Compare average session duration and pages per session by device category and browser over the last 90 days

This demonstrates:

• Multi-dimensional analysis
• Time series comparison
• User engagement metrics
• Technology segmentation

3. Traffic Source Performance

Show me conversion rates and revenue by traffic source and campaign, comparing last 30 days vs previous 30 days

This demonstrates:

• Marketing performance analysis
• Period-over-period comparison
• Conversion tracking
• Revenue attribution

4. Content Performance

What are my top 10 pages by engagement rate, and how has their performance changed over the last 3 months?

This demonstrates:

• Content analysis
• Trend analysis
• Engagement metrics
• Ranking and sorting

---

🚀 Performance Optimizations

This MCP server includes built-in optimizations to prevent context window crashes and ensure smooth operation:

Smart Data Volume Management

• Automatic row estimation - Checks data volume before fetching
• Interactive warnings - Alerts when queries would return >2,500 rows
• Optimization suggestions - Provides specific recommendations to reduce data volume

Server-Side Processing

• Intelligent aggregation - Automatically aggregates data when beneficial (e.g., totals across time periods)
• Smart sorting - Returns most relevant data first (recent dates, highest values)
• Efficient filtering - Leverages GA4's server-side filtering capabilities

User Control Parameters

• limit - Set maximum number of rows to return
• proceed_with_large_dataset=True - Override warnings for large datasets
• enable_aggregation=False - Disable automatic aggregation
• estimate_only=True - Get row count estimates without fetching data

Example: Handling Large Datasets

# This query would normally return 2,605 rows and crash context window
get_ga4_data(
    dimensions=["date", "pagePath", "country"],
    date_range_start="90daysAgo"
)
# Returns: {"warning": True, "estimated_rows": 2605, "suggestions": [...]}

# Use monthly aggregation instead
get_ga4_data(
    dimensions=["month", "pagePath", "country"], 
    date_range_start="90daysAgo"
)
# Returns: Clean monthly data with manageable row count

---

Available Tools

The server provides a suite of tools for data reporting and schema discovery.

1. search_schema - Searches for a keyword across all available dimensions and metrics. This is the most efficient way to discover fields for a query.
2. get_ga4_data - Retrieve GA4 data with built-in intelligence for better and safer results (includes data volume protection, smart aggregation, and intelligent sorting).
3. list_dimension_categories - Lists all available dimension categories.
4. list_metric_categories - Lists all available metric categories.
5. get_dimensions_by_category - Gets all dimensions for a specific category.
6. get_metrics_by_category - Gets all metrics for a specific category.
7. get_property_schema - Returns the complete schema for the property (Warning: this can be a very large object).

---

Dimensions & Metrics

Access to 200+ GA4 dimensions and metrics organized by category:

Dimension Categories

• Time: date, hour, month, year, etc.
• Geography: country, city, region
• Technology: browser, device, operating system
• Traffic Source: campaign, source, medium, channel groups
• Content: page paths, titles, content groups
• E-commerce: item details, transaction info
• User Demographics: age, gender, language
• Google Ads: campaign, ad group, keyword data
• And 10+ more categories

Metric Categories

• User Metrics: totalUsers, newUsers, activeUsers
• Session Metrics: sessions, bounceRate, engagementRate
• E-commerce: totalRevenue, transactions, conversions
• Events: eventCount, conversions, event values
• Advertising: adRevenue, returnOnAdSpend
• And more specialized metrics

---

Troubleshooting

If ga4-mcp-server is not found:

• Use the explicit interpreter launch style instead: python -m ga4_mcp
• Reinstall with the same Python interpreter your MCP client will use

If you get No module named ga4_mcp:

/full/path/to/python -m pip install google-analytics-mcp

Install the package with the exact interpreter you reference in your MCP configuration.

Permission errors:

# Try user install instead of system-wide
python -m pip install --user google-analytics-mcp

If the server says the credentials file is missing:

1. Verify the JSON file path is absolute, correct, and accessible
2. Check service account permissions:
   • Go to Google Cloud Console → IAM & Admin → IAM
   • Find your service account → Check permissions
3. Verify GA4 access:
   • GA4 → Admin → Property access management
   • Check for your service account email

If the server says GA4_PROPERTY_ID is invalid or queries return no data:

• Use the numeric Property ID (for example 123456789)
• Do not use the Measurement ID (for example G-XXXXXXXXXX)
• Confirm the service account has at least Viewer access on that property

API quota/rate limit errors:

• GA4 has daily quotas and rate limits
• Try reducing the date range in your queries
• Wait a few minutes between large requests

---

Project Structure

google-analytics-mcp/
├── ga4_mcp/                # Main package directory
│   ├── server.py           # Core server logic
│   ├── coordinator.py      # MCP instance
│   └── tools/              # Tool definitions (reporting, metadata)
├── pyproject.toml          # Package configuration for PyPI
├── requirements.txt        # Dependencies for local dev
├── README.md               # This file
└── ...

---

License

Apache License 2.0