Mcp Server Gcp Manager

@KishoreKu

4 months ago

Overview

**Google Cloud MCP Server

An open-source Model Context Protocol (MCP) server that enables AI agents (like Claude, Llama 3.1, or custom LangChain bots) to interact with Google Cloud Platform safely and autonomously.

This server acts as a bridge between your LLM and the gcloud CLI, allowing you to manage Cloud Run services and Secrets using natural language.

🚀 Features

🔍 List Services: detailed inspection of running Cloud Run services.
🔐 Manage Secrets: Safely create and update secrets in Google Secret Manager.
- Self-Healing: Automatically detects if the Secret Manager API is disabled and enables it for you.
🚀 Deploy & Update: Deploy new revisions or update memory/image configuration for Cloud Run services.
- Safe-Guards: Checks existing configuration before applying changes to prevent accidental overwrites.

🛠️ Prerequisites

Python 3.10+
Google Cloud SDK (gcloud): Must be installed and authenticated.
- Install: brew install --cask google-cloud-sdk (Mac) or see Official Docs.
- Auth: Run gcloud auth login and gcloud config set project [YOUR_PROJECT_ID].

📦 Installation

Clone the repository:

git clone https://github.com/YOUR_USERNAME/mcp-gcp-manager.git
cd mcp-gcp-manager

Create a Virtual Environment (Recommended):

python3 -m venv venv
source venv/bin/activate

Install Dependencies:
```
pip install mcp
```

⚙️ Configuration

The server attempts to automatically find your gcloud executable.

If you see a "gcloud not found" error: You can manually tell the server where gcloud is by setting an environment variable:

# 1. Example for Mac/Linux
export GCLOUD_PATH="/usr/local/bin/gcloud"

# 2. Example for Windows (PowerShell)
$env:GCLOUD_PATH="C:\Program Files\Google\Cloud SDK\google-cloud-sdk\bin\gcloud.exe"

💻 Usage

Option 1: Using with Claude Desktop To use this with the official Claude Desktop app, add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "gcp-manager": {
      "command": "/path/to/your/venv/bin/python",
      "args": ["/path/to/mcp-gcp-manager/server.py"]
    }
  }
}

Option 2: Using with LangChain / Custom Python Agents

You can connect to this server using langchain-mcp-adapters:

from langchain_mcp_adapters.client import MultiServerMCPClient

# 3. Connect to the local server process
client = MultiServerMCPClient({
    "gcp": {
        "command": "python",
        "args": ["server.py"],
        "transport": "stdio"
    }
})

# 4. Get tools and bind to your model (Llama 3, GPT-4, etc.)
tools = await client.get_tools()

🛡️ Tools Available

Tool Name	Description
`list_running_services`	Lists all Cloud Run services in a specific project.
`create_secret`	Creates/Updates secrets. Handles API enablement automatically.
`deploy_service`	Deploys a new revision. Supports updating memory, image, etc.

🤝 Contributing

Pull requests are welcome! Please ensure any new tools include proper error handling and do not hardcode local paths.

📄 License

MIT**

Try in Playground

Server Config

{
  "mcpServers": {
    "gcp-manager": {
      "command": "python",
      "args": [
        "/path/to/mcp-server-gcp-manager/server.py"
      ],
      "env": {
        "GCLOUD_PATH": "/path/to/your/gcloud"
      }
    }
  }
}

Build with ShipAny.