Gemini MCP Server

A streamlined Model Context Protocol (MCP) server that provides Gemini API documentation lookup and ready-to-use code snippets. This server is built using FastMCP and is designed to be easily hosted locally or on a remote VM (e.g., Oracle Cloud).

Features

• search_library_docs: Retrieves official documentation URLs for various Gemini topics (e.g., caching, function calling, tokens).
• get_gemini_code_snippet: Provides copy-pasteable Python code snippets for common Gemini tasks (e.g., chat, file upload, list models).
• Secure Authentication: Uses a secret token to authenticate requests from your MCP client.

Prerequisites

• Python 3.10 or higher
• pip or uv (recommended for faster dependency management)

Installation

1. Clone the repository:

   git clone https://github.com/SOBANEJAZ/Gemini-MCP-Server.git
   cd Gemini-MCP-Server
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Configure Environment Variables: Create a .env file in the root directory. You can copy the template below:

   cp .env.example .env  # If provided, otherwise create new
   nano .env
   

   Example .env content:

   # Security token for the API header (Client -> Server)
   # REPLACE THIS with a strong, random string
   MCP_SECRET=my_super_secret_token_123
   
   # Server bind address and port
   HOST=0.0.0.0
   PORT=8000
   

Running Locally

To start the server locally:

python server.py

You should see output indicating the server is running: Starting Gemini MCP Server on 0.0.0.0:8000...

Hosting on a VM (e.g., Oracle Cloud)

1. Deploy the code: Clone the repo and install dependencies on your VM as described in the Installation section.

2. Firewall Configuration: Ensure your VM's firewall allows ingress traffic on port 8000.

   • Oracle Cloud: Go to your VCN > Security Lists > Ingress Rules. Add a rule for Source CIDR 0.0.0.0/0, Protocol TCP, and Destination Port Range 8000.
   • Ubuntu Firewall (if enabled):

     sudo ufw allow 8000/tcp
     

3. Run in Background: Use nohup to keep the server running after you disconnect:

   nohup python server.py > server.log 2>&1 &
   

   Or use a process manager like supervisor or systemd for production.

MCP Client Configuration (Claude Code/Cursor/Antigravity)

Add the server to your mcp_config.json. Choose the option that matches your setup.

Option A: Local Connection

If the server is running on the same machine as your client.

{
  "mcpServers": {
    "Gemini-MCP-Server": {
      "command": "python",
      "args": [
        "/absolute/path/to/gemini-mcp-server/server.py"
      ],
      "env": {
        "MCP_SECRET": "my_super_secret_token_123"
      }
    }
  }
}

Option B: Remote Connection (SSE over HTTP)

If the server is on a VM or different machine. This uses the mcp-remote tool (install via npm install -g mcp-remote or use npx).

{
  "mcpServers": {
    "Gemini-MCP-Server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://<YOUR_VM_IP>:8000/sse",
        "--allow-http",
        "--insecure",
        "--header",
        "API: <YOUR_MCP_SECRET>"
      ]
    }
  }
}

Option C: HTTPS / Smithery Connection

If you are hosting behind a platform that handles SSL termination (like Smithery, Cloudflare, or an Nginx reverse proxy), the server.py continues to run on HTTP while the platform handles HTTPS externally.

{
  "mcpServers": {
    "Gemini-MCP-Server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://<YOUR_SECURE_DOMAIN>/sse",
        "--header",
        "API: <YOUR_MCP_SECRET>"
      ]
    }
  }
}

Note: Replace <YOUR_VM_IP> or <YOUR_SECURE_DOMAIN> with your server's address, and <YOUR_MCP_SECRET> with the secret you set in your .env file.