Google Image Search MCP

A Python-based MCP (Model Context Protocol) server that provides tools for searching and downloading images using Google Image Search.

Features

• Search for images using Google Image Search API
• Download images to local storage

Installation

1. Clone the repository
2. Install uv _(just once, if needed)_
3. Create a virtual environment: _(just once)_

   uv venv

1. Activate the virtual environment:

Linux/macOS:

source .venv/bin/activate

Windows:

.venv\Scripts\activate

1. Install dependencies:

Linux/macOS:

uv pip install -r requirements.txt

Windows:

uv pip install -r requirements.txt

1. Create a .env file with your SerpAPI key:

(Get your key here: https://serpapi.com/dashboard)

   SERP_API_KEY=your_api_key_here

Usage

You can either run the server:

uv run main.py

or test the MCP server with inspector:

npx @modelcontextprotocol/inspector uv run main.py

Run the server:

uv run main.py

In VSCode, add this MCP server configuration to your mcp-servers.json file:

{
  "servers": {
    "search-images": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\Users\\YOURUSERNAME\\google-image-search-mcp-python",
        "run",
        "main.py"
      ]
    }
  }
}

The server provides the following tools:

1. search_images_tool: Search for images using Google Image Search

• Parameters:
• query: The search query for finding images
• limit: Maximum number of results to return (default: 10)

1. download_image_tool: Download an image to a local directory

• Parameters:
• image_url: URL of the image to download
• output_path: Directory path where the image should be saved
• filename: Filename for the downloaded image (including extension)

Examples

Prompt:

use your search images tool to search for 5 images about "cute cats" and download the best one in ./gatitos/

Code

# Search for images
results = await search_images_tool("cute puppies", limit=5)

# Download an image
saved_path = await download_image_tool(
    image_url="https://example.com/image.jpg",
    output_path="./images",
    filename="puppy.jpg"
)

Error Handling

The tools provide detailed error messages when something goes wrong. All errors are logged to stderr and returned in a structured format with an isError flag.

Dependencies

• mcp-server: For MCP server functionality
• python-dotenv: For environment variable management
• aiohttp: For async HTTP requests