gemini-cloud-assist

The Gemini Cloud Assist MCP server bridges MCP clients with the Gemini Cloud Assist APIs, enabling you to manage and troubleshoot Google Cloud environments using natural language prompts from tools like the Gemini CLI. With this MCP server, you can create and run investigations, add observations to refine analyses, and fetch investigation details. The Gemini CLI can be configured to automatically pull and run this MCP server on demand via npx, giving you a seamless on-demand experience without permanently installing the server locally. Typical workflows include starting an investigation into a cloud issue, running analytical steps to gather data, and iteratively refining the investigation by adding observations or querying results.

Key MCP tools exposed by this server include: create_investigation (to initialize a new Cloud Assist investigation), run_investigation (to execute analysis against a given investigation), add_observation (to append insights or findings to an investigation), and fetch_investigation (to retrieve investigation details or list investigations). These tools work through the Gemini CLI and the MCP protocol, enabling interactive troubleshooting and root-cause analysis directly from your terminal.

Prerequisites:

• Node.js v20 or later and npm
• Git
• Google Cloud SDK (gcloud) installed and configured
• A Google Cloud project with appropriate IAM roles:
  • roles/serviceusage.serviceUsageAdmin to enable Cloud Assist APIs
  • roles/geminicloudassist.user to access Cloud Assist APIs

Step 1: Install Node.js and set up your environment

• Install Node.js from https://nodejs.org
• Ensure npm is available: npm -v

Step 2: Authenticate with Google Cloud (ADC)

• Authenticate your user account to gcloud: gcloud auth login
• Set up Application Default Credentials for the server: gcloud auth application-default login

Step 3: Run the MCP server on demand via npx

• This MCP server is published as an npm package. Use npx to pull and run it when needed:

# Run the MCP server on demand using npx (no permanent installation required)
npx -y @google-cloud/gemini-cloud-assist-mcp@latest

Step 4: Optional – configure an MCP client

• You can embed the following MCP config in your Gemini CLI settings to use the on-demand server:

"mcpServers" : {
  "GeminiCloudAssist": {
    "command": "npx",
    "args": ["-y", "@google-cloud/gemini-cloud-assist-mcp@latest"],
    "timeout": 600000
  }
}

Step 5: Validate installation

• Ensure the MCP server is reachable by your MCP client and that authentication to Google Cloud via ADC works. Run a basic query like listing investigations to confirm connectivity.

Environment and configuration tips:

• ADC (Application Default Credentials) is used for Google Cloud API authentication. Ensure gcloud auth and gcloud auth application-default login have been completed as described in the Quick Start.
• The MCP server relies on the Gemini Cloud Assist APIs; ensure your Google Cloud project has the required Cloud Assist APIs enabled and that your account has the listed IAM roles.
• If you prefer a persistent installation, you can install the MCP server package globally or as a project dependency, but the recommended approach in the docs is to run via npx for on-demand usage.
• When integrating with Gemini CLI, you can install the MCP server as a Gemini extension for a more integrated experience or configure project/global MCP settings to reuse the server across sessions.