home / mcp / imagesorcery mcp server

ImageSorcery MCP Server

An MCP server providing tools for image processing operations

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "sunriseapps-imagesorcery-mcp": {
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}

ImageSorcery MCP enables your AI assistant to perform advanced local image processing tasks using a collection of tools powered by computer vision models. Run everything locally to crop, resize, annotate, detect objects, extract text, and more without sending images to external servers.

How to use

Connect an MCP client to the ImageSorcery MCP server and start issuing image-processing commands. You can combine multiple tools to accomplish complex tasks such as detecting objects in an image, cropping regions, drawing annotations, removing backgrounds, or applying watermarks. Ensure you reference the local server endpoint so your client can relay commands and receive results.

How to install

Prerequisites: you need Python 3.10 or higher, and either pipx or a virtual environment setup with Python. You also need system libraries required by OpenCV (ffmpeg, libsm6, libxext6, libgl1-mesa-glx). If you plan to run the server locally, you should have a MCP client ready (e.g., Claude.app, Cline, or another MCP client).

Option 1 — Using pipx (recommended)

pipx install imagesorcery-mcp
imagesorcery-mcp --post-install

Option 2 — Manual virtual environment (Plan B)

python -m venv imagesorcery-mcp
source imagesorcery-mcp/bin/activate  # Linux/macOS
# Windows: imagesorcery-mcp\Scripts\activate
pip install imagesorcery-mcp
imagesorcery-mcp --post-install

Tip: If you use a persistent environment, you will reference the executable path in your MCP client configuration, for example "/full/path/to/venv/bin/imagesorcery-mcp". For Windows, use the corresponding Scripts path.

Running the server

You can run ImageSorcery MCP in different modes. The standard local mode is STDIO and uses the default command: imagesorcery-mcp.

imagesorcery-mcp

If you prefer web-based deployments, you can use Streamable HTTP mode with a configurable host, port, and path, or use HTTP mode with an explicit endpoint. The HTTP mode endpoint example is shown below. Use the default STDIO mode for local development and testing.

imagesorcery-mcp --transport=streamable-http
# Or with custom host/port/path
imagesorcery-mcp --transport=streamable-http --host=0.0.0.0 --port=4200 --path=/mcp

# HTTP mode example endpoint (client connects here)
http://127.0.0.1:8000/mcp

Available tools

blur

Blurs specified areas of an image using OpenCV; supports rectangular or polygonal regions and can blur the background by inverting the area.

change_color

Changes the color palette of an image, enabling effects like sepia or grayscale.

config

View or update MCP configuration settings at runtime.

crop

Crop an image using OpenCV numpy slicing based on coordinates.

detect

Detect objects in an image using Ultralytics models; supports return of segmentation masks or polygons with a confidence threshold.

draw_arrows

Draw arrows on an image to indicate directions or highlights.

draw_circles

Draw circles on an image at specified centers and radii.

draw_lines

Draw lines on an image between two points with color and thickness.

draw_rectangles

Draw one or more rectangles on an image, with options for filled or outlined shapes.

draw_texts

Overlay text onto an image at specified positions.

fill

Fill areas of an image with a color or transparency; can also invert areas or remove backgrounds.

find

Find objects in an image based on a text description; can return segmentation masks or polygons.

get_metainfo

Retrieve metadata information for an image file.

ocr

Perform Optical Character Recognition on images to extract text using OCR models.

overlay

Overlay one image onto another with proper handling of transparency.

resize

Resize an image to specified dimensions.

rotate

Rotate an image by a given angle while preserving content.