Quick Start¶

Prerequisites¶

• Node.js 18+ — for the MCP server itself
• gcloud CLI — authenticated with an account that has appropriate GCA admin roles
• An AI assistant that supports MCP (Claude Code, Gemini CLI, Cursor, etc.)

Option 1 — Interactive setup (recommended)¶

Run the interactive installer. It auto-detects which AI clients are installed and patches their config:

npx gemini-code-assist-admin-mcp

You'll see a menu like:

? Where would you like to add the MCP server? (Use arrow keys)
❯ Claude Code (claude_desktop_config.json)
  Gemini CLI (settings.json)
  Antigravity / agy
  Manual (show config snippet)

Option 2 — Global install¶

If you prefer a persistent install with the short gca-admin alias:

npm install -g gemini-code-assist-admin-mcp
gca-admin

Option 2 — Client-specific guides¶

Pick your client:

• Claude Code
• Gemini CLI
• Antigravity (agy)
• Cursor / Windsurf
• Any MCP client

Verify it's working¶

Once installed, ask your assistant:

"Is Gemini Code Assist enabled in my current GCP project?"

The assistant should call check_gca_status and return the API enablement status.

Authentication¶

The MCP server inherits your local gcloud credentials. Make sure you're authenticated:

gcloud auth login
gcloud auth application-default login
gcloud config set project YOUR_PROJECT_ID