Google × Claude MCP Integration | Connecting Drive, GA4, and GSC

What Is Google Claude MCP? — Two Ways to Connect Claude to Google

MCP is an open standard published by Anthropic that provides a common protocol for connecting conversational models like Claude to external services and data sources. There are two main paths for connecting Claude to Google services:

1. Standard Connectors: Pre-verified built-in connections from Anthropic. Google Drive, Gmail, and Google Calendar are supported, and can be enabled in just a few clicks from the settings screen.
2. MCP Servers: For services not covered by standard connectors (GA4, Search Console, Ads, BigQuery, etc.), you set up your own MCP server to establish the connection.

In practice, the split is simple: use standard connectors for everyday Workspace tools, and MCP servers for data analytics and marketing platforms.

Connecting Workspace Apps via Standard Connectors (Drive, Gmail, Calendar)

Google Workspace connectors (Gmail / Google Drive / Google Calendar) are available to all Claude and Claude Desktop users. In Anthropic's directory, Drive ranks 1st, Gmail 2nd, and Calendar 3rd — they are among the most popular connectors available.

The steps to add them are as follows:

1. Open "Connectors" from Claude's settings.
2. Select Google Drive, Gmail, or Google Calendar from the list and click "Connect."
3. Complete OAuth authentication (grant access) with your Google account.
4. Once connected, you can give natural-language instructions in chat, such as "Find last week's invoice emails."

Note that these built-in connectors currently focus on searching, reading, and creating — they do not support update operations like moving, renaming, or deleting existing files. For Team and Enterprise plans, the organization Owner (admin) must first enable connectors at the organization level. As of March 31, 2026, Skills, Connectors, and Plugins have been consolidated into claude.ai/directory.

Connecting Data Services via MCP Servers (GA4, Search Console, Ads)

Data-oriented services such as Google Analytics (GA4), Search Console, Google Ads, and BigQuery are not included in the standard connectors, so you need to set up an MCP server separately. Common options include:

• Google Analytics (GA4): The official Google MCP server google-analytics-mcp (repository: googleanalytics/google-analytics-mcp). Lets you extract traffic insights and generate improvement suggestions using natural language.
• Google Search Console: An MCP server for GSC. Retrieves search rankings, impressions, and inbound queries to automatically generate SEO reports.
• Google Ads: Available via MCP servers provided by vendors such as cdata, giving you natural-language access to advertising data.
• Google Search: Some servers, like mcp-gemini-google-search, can be launched with npx.

These are typically used in combination with Claude Code or Claude Desktop.

Adding a Google MCP Server in Claude Code

In Claude Code, you register a server using the claude mcp add command. Choose from three scopes: local, project, or user.

1. Run claude mcp add in your terminal and specify the server name, startup command, and scope. --scope project shares the configuration with your team via .mcp.json; --scope local applies only to your environment; --scope user applies across all projects. Here is an example of registering the GA4 MCP server with project scope:

   # project scope (written to .mcp.json and shared with the team)
   claude mcp add ga4 --scope project -- uv run analytics-mcp
   

2. To share settings across a project, write the configuration to .mcp.json.

3. Pass credentials (such as a service account key) via environment variables.

4. Start claude and verify that the registered MCP tools are loaded.

Here is an example of registering GA4 in .mcp.json. Note that uv specified in command is a fast Python package manager that must be installed in advance (if you see command not found, install uv first). Installing the repository google-analytics-mcp introduced in the previous section makes the analytics-mcp command available to run with uv run.

{
  "mcpServers": {
    "ga4": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "analytics-mcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account-key.json",
        "GOOGLE_CLOUD_PROJECT": "your-project-id"
      }
    }
  }
}

Setting Up MCP in Claude Desktop

In Claude Desktop, you describe MCP server information in JSON inside the configuration file claude_desktop_config.json.

1. From Claude Desktop settings, go to "Developer" → "Edit Configuration File."
2. Add the target server's command, args, and any required environment variables under mcpServers.
3. Restart Claude Desktop to load the settings.
4. Once connected, you can freely query your Google service data.

For servers launched with npx, simply specify npx as the command and the package name in args.

Authentication Methods — OAuth vs. Service Account

One of the most common stumbling blocks with Google integrations is the difference in authentication methods. Here is how to choose:

• OAuth: Used by standard connectors (Drive / Gmail / Calendar) and Remote MCP. This is the user account authorization method — you simply approve access in your browser.
• Service Account: Commonly used for API-based MCP integrations like GA4. You create a service account in Google Cloud, generate a JSON key, and pass it via GOOGLE_APPLICATION_CREDENTIALS.

A good rule of thumb: use OAuth for personal use, and a service account for server-resident setups or automated data analysis. For more on Remote MCP, see How to Use Claude Remote MCP.

Compatibility by Product — Claude.ai / Code / Desktop / Cowork

For more detail on how to use each product, reading Claude Cowork × Google Integration and Google Search Console and Claude Integration together will deepen your understanding.

Common Issues and Solutions

• Write operations don't work: The standard Drive connector focuses on reading and creating — it cannot move or delete files. If you need full write automation, consider an MCP server or a full Cowork integration.
• Connectors not visible in Claude Code: Even if you have connected through claude.ai, the connection may not carry over to a Claude Code session. Re-registering as an MCP server in Code is the most reliable approach.
• Cannot connect in Team / Enterprise: The organization's Owner may not have enabled connectors in the admin panel. Contact your admin to have them activated.
• Authentication errors: Common causes include an incorrect path to the service account key or missing permissions for the target GA4 property.

Summary

The answer to "google claude mcp" lies in using two paths: standard connectors for Workspace apps, and MCP servers for data services. If you remember that Drive, Gmail, and Calendar connect via a quick OAuth flow, while GA4, Search Console, and Ads require registering a server with claude mcp add or a config file, you have everything you need to use Claude as a practical hub for your Google data.

Sources

• Use Google Workspace connectors | Claude Help Center
• Getting Started with Claude MCP Configuration | Practical Steps Including Google Search Integration | Murasan Lab
• Trying Out MCP Server Integration Between Claude Code & Google Analytics (GA4) | Zenn
• MCP: Google Ads × Claude Desktop Integration Guide | CData