Creativity · MCP — server

MCP Google Cloud Server

Google Cloud's MCP support is delivered as a collection of targeted servers — Vertex AI agent-builder tools, BigQuery, Cloud Run, Cloud Logging, and a general `gcloud` wrapper. Most are published under the GoogleCloudPlatform GitHub org with community packages filling gaps. They authenticate via Application Default Credentials (ADC), so the MCP server inherits whatever IAM policy your service account carries. This page is the GCP-landing overview.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python / Go
Transports: stdio, http

Capabilities

Servers for BigQuery, Vertex AI, Cloud Run, Cloud Logging, Firestore, GCS
Resources: project/region metadata; per-service entity URIs
Auth: Application Default Credentials (ADC) via service-account key or gcloud

Install

uvx mcp-server-gcp

Configuration

{
  "mcpServers": {
    "gcp": {
      "command": "uvx",
      "args": ["mcp-server-gcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/Users/you/.gcp/mcp-sa.json",
        "GCP_PROJECT": "my-project"
      }
    }
  }
}

Frequently asked questions

Which Google Cloud services have MCP servers?

Published or in-preview servers cover BigQuery, Vertex AI, Cloud Run, Cloud Logging, Firestore, and GCS. Check github.com/GoogleCloudPlatform for the current catalogue.

How do I authenticate?

Set GOOGLE_APPLICATION_CREDENTIALS to a service-account key JSON or rely on gcloud auth application-default login for local dev. Use Workload Identity Federation when running in another cloud.

Can Gemini agents consume these servers?

Yes. Google's Gemini-based agent tooling interoperates with MCP, so you can reuse the same servers across Claude, Gemini, and OpenAI agents.

Sources

Google Cloud MCP — accessed 2026-04-20
GoogleCloudPlatform GitHub org — accessed 2026-04-20