Google Cloud Joins MCP: How to Connect Claude Code to BigQuery

Key Takeaways

• Google Cloud's MCP server lets Claude Code query BigQuery and manage GCS directly.
• Install it with claude mcp add google-cloud and authenticate.

What Changed

Google Cloud has adopted the Model Context Protocol (MCP) natively, making it the first major cloud provider to ship an official MCP server for its platform. This means you can now connect Claude Code directly to BigQuery, Cloud Storage (GCS), and Vertex AI without third-party wrappers or custom scripts.

The MCP ecosystem has exploded to over 13,000 servers as of late June 2026, but quality varies wildly — 54% of those have zero community adoption. Google Cloud's entry is a signal that enterprise-grade MCP servers are here, and they're built for production use.

What It Means For You

If you're a Claude Code user working in GCP, this changes your daily workflow. Instead of context-switching to the GCP Console or running separate gcloud commands, you can:

• Query BigQuery tables directly from Claude Code: "Show me the top 10 customers by revenue from the last quarter"
• List and manage Cloud Storage buckets: "Upload this file to my data-lake bucket"
• Invoke Vertex AI models: "Run a prediction using my deployed model"

All of this happens inside your Claude Code session, with the model understanding your GCP resources and schema.

Try It Now

Installation

claude mcp add google-cloud

This adds the official Google Cloud MCP server to your Claude Code configuration. You'll be prompted to authenticate with your GCP credentials — the server uses Application Default Credentials (ADC), so if you already have gcloud configured, it should work seamlessly.

Configure Access

By default, the server has access to all projects your credentials can reach. To scope it down (recommended for production), edit your ~/.claude/settings.json:

{
  "mcpServers": {
    "google-cloud": {
      "command": "npx",
      "args": ["@google-cloud/mcp-server"],
      "env": {
        "GOOGLE_CLOUD_PROJECT": "my-project-id",
        "MCP_ALLOWED_SERVICES": "bigquery,storage"
      }
    }
  }
}

Example Workflow

Start a Claude Code session and try:

> List my BigQuery datasets in project my-project-id
> What tables are in the `analytics` dataset?
> Write a SQL query to find the top 10 products by sales in the `orders` table and execute it

Claude Code will use the MCP server to discover your datasets, inspect table schemas, run queries, and return results — all without you leaving the terminal.

Why This Matters

Google's adoption of MCP is significant. They've invested heavily in their own agent frameworks (ADK Go 2.0, released last week with graph-based workflows and human-in-the-loop). Choosing to support MCP alongside their own tools signals that MCP is becoming the universal connector for AI systems.

It also means Claude Code users get first-class access to GCP without waiting for community-built servers that may or may not be maintained. Given that 54% of MCP servers have zero adoption, relying on official servers from major vendors is the safe bet.

Limitations

• Read-only by default: The server starts in read-only mode for BigQuery and Storage. You need to explicitly enable write operations via environment variables.
• Vertex AI is limited: Currently supports model invocation but not training or deployment workflows.
• No IAM management: You can't modify permissions through the MCP server — use the GCP Console or gcloud for that.

Source: news.google.com