GCP MCP Setup

OneMCP Setup Skill

Overview

This skill provides the automated steps to configure your environment and enable the Core Google Managed MCP servers (Logging, Monitoring, GKE, Run, Resource Manager, Developer Knowledge) by default. Use the --all flag to include additional services like Error Reporting, Cloud SQL, Spanner, Pub/Sub, Vertex AI, Firestore, and BigQuery.

When to Use

Use this skill when the user asks to "install OneMCP", "setup OneMCP", or "configure Google Managed MCP servers".

Core Pattern

1. Prerequisites Verification: Ensure the user has gcloud authenticated and a target Google Cloud Project with billing enabled.
2. Setup Script Execution: Run the provided script to enable APIs, MCP services, and generate the required API Key.
   • By default, it installs a Lean SRE toolset (Logging, Monitoring, GKE, Run, Resource Manager, Developer Knowledge).
   • Use the --all flag to include optional services (Error Reporting, Databases, Vertex AI).
   • Use --harness to target a specific CLI harness (gemini, antigravity, or copilot). Defaults to gemini+antigravity.
   • Script: skills/gcp-mcp-setup/scripts/setup_onemcp.py <PROJECT_ID> [--local | --global] [--harness {gemini,antigravity,copilot}] [--all] [--google-maps-key YOUR_KEY]
3. Configure Settings: The script automatically writes the mcpServers configuration to the correct location depending on the target harness:
   • Gemini CLI: .gemini/settings.json (local) or ~/.gemini/settings.json (global).
   • Antigravity CLI (agy): .gemini/antigravity/mcp_config.json and .gemini/config/mcp_config.json (local) or the global equivalents under ~/.gemini/.
   • GitHub Copilot CLI: ~/.copilot/mcp-config.json (always global — no workspace-level config yet).
4. Verification & Diagnostics:
   • Script: skills/gcp-mcp-setup/scripts/verify_setup.py [--harness {gemini,antigravity,copilot}]
   • This script checks identity consistency (gcloud vs ADC), runs the harness-specific MCP list command, and confirms all expected google-* servers are registered.

Examples

User: "Please install OneMCP on my project my-gcp-project" Agent: "I will use the gcp-mcp-setup skill to configure your project and Gemini settings for Cloud Logging, Developer Knowledge, Firestore, and BigQuery."

Authentication

Note that the gcloud CLI Identity (how you are currently logged in via gcloud auth login) may differ from the MCP Server Identity (which is based on Application Default Credentials, or ADC).

MCP servers typically use ADC to authenticate. If these identities do not match, you may encounter permission errors (e.g., serviceusage.serviceUsageConsumer denied) even if gcloud commands work correctly.

If MCP doesn't work, you can propose to run this on user's behalf: gcloud auth application-default login --account <PROPOSED_IDENTITY_EMAIL>

💡 Path Forward

If you encounter an identity mismatch:

1. Run gcloud auth application-default login --account=YOUR_CLI_IDENTITY to synchronize identities.
2. Verify the identities match by running the verification script: python3 skills/gcp-mcp-setup/scripts/verify_setup.py.

Documentation

• Antigravity MCP integration guide: https://antigravity.google/docs/mcp
• Code samples from Romin Irani: https://github.com/rominirani/google-mcp-servers/blob/main/demos/README.md
• All supported products: https://docs.cloud.google.com/mcp/supported-products