The Image MCP

🖼️ the-image-mcp

A simple MCP server for AI models to find, preview, download, and edit pictures from the web.
Powered by Chrome DevTools Protocol (CDP) and Pillow. Zero Playwright bloat.

⚡ What does it do?

1. 🔍 Search & Preview

Searches the web (Google Images with automatic Bing fallback) and hands low-res thumbnails to the AI so it can inspect candidate images, compare options, and pick the winner.

2. 🚀 Grab Full Resolution

Takes the chosen candidate's URL and downloads the crystal-clear, uncompressed original picture (up to 4K+) directly to disk.

3. 🛠️ Edit & Transform

Tool	What it does
resize_image	Resize by width or height while keeping aspect ratio
crop_image	Cut out any exact bounding box (left, top, right, bottom)
convert_image_format	Convert between PNG, JPEG, WEBP, BMP, TIFF
rotate_flip_image	Rotate by any angle or mirror horizontally / vertically
apply_image_filter	Apply blur, grayscale, sharpen, contour, or edge_enhance
get_image_metadata	Inspect dimensions, DPI, color mode, transparency, and file size

---

🚀 Quick Setup

1. Install & Run

Make sure you have uv installed:

# Clone and enter the project
git clone https://github.com/phoenix-xr/the-image-mcp.git
cd the-image-mcp

# Sync dependencies
uv sync

# Start the server
uv run images-mcp

---

2. Connect to Your AI (Claude, Antigravity, Cursor)

Add this block to your mcp_config.json or claude_desktop_config.json:

{
  "mcpServers": {
    "the-image-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\Users\\anshy\\Documents\\project-gits\\images-mcp",
        "run",
        "images-mcp"
      ]
    }
  }
}

---