gcp

This MCP server provides a bridge between GenAI applications and Google Cloud Platform (GCP) services. It exposes a suite of tools that allow you to manage GCP resources from an MCP-enabled assistant, including Compute Engine (VMs), Cloud Storage, Cloud Functions, BigQuery, Cloud SQL, GKE, Cloud Run, Pub/Sub, IAM, Cloud Build, Vertex AI, and monitoring/logging capabilities. You can invoke these tools via MCP tool payloads (JSON objects) that specify the tool name and its arguments. The server handles authentication through various methods (service account keys, ADC, OAuth, and more), and supports multi-project contexts for orchestration across environments.

Prerequisites:

• Python 3.8+ installed on your machine
• pip (Python package installer) available
• Access to a GCP service account key if not using ADC

Installation steps:

1. Install the MCP server package from PyPI:

   pip install gcp-mcp-server
   

2. (Optional) Install from source if you prefer the development version:

   git clone https://github.com/LokiMCPUniverse/gcp-mcp-server.git
   cd gcp-mcp-server
   pip install -e .
   

3. Verify installation by importing in Python (quick sanity check):

   python -c "from gcp_mcp import GCPMCPServer; print(GCPMCPServer)"
   

4. Configure your environment variables as needed (see the mcp_config for defaults) and prepare credentials as described in the README (GOOGLE_APPLICATION_CREDENTIALS or ADC).

Notes and tips:

• Use a dedicated service account with the minimum required permissions for the GCP resources you intend to manage.
• When using multiple projects, configure the corresponding project IDs and credentials (GCP_PROD_, GCP_DEV_) as needed.
• If you encounter authentication issues, ensure GOOGLE_APPLICATION_CREDENTIALS points to a valid JSON key file or rely on Application Default Credentials in your environment.
• The MCP server supports a variety of tools; consult the README for specific tool names and required arguments. You can extend or customize tools as needed by your implementation.
• The environment can be adjusted per deployment; avoid exposing sensitive credentials in logs or config payloads.