imagesorcery

ImageSorcery MCP exposes a Python-based MCP server that runs locally to provide computer-vision backed image processing tools to your AI assistant. It integrates a suite of vision capabilities such as cropping, resizing, background removal, text extraction, object detection, drawing shapes or texts on images, applying filters, and model-based analysis via a unified MCP interface. You can use prompts or tool invocations from an MCP client (e.g., Claude, Cline, Cursor) to apply these operations to your local image assets without sending data to external servers. The tools are designed to be combined into multi-step workflows (e.g., detect objects in an image, then crop and annotate regions, then export results) using the provided tool prompts and the ImageSorcery resources. The server also exposes a config tool to view or adjust settings like detection confidence and telemetry, enabling customization for your environment.

Prerequisites

• Python 3.10 or higher is recommended
• pipx (recommended) for virtual environments and easy installation
• System libraries required by OpenCV (ffmpeg, libsm6, libxext6, libgl1-mesa-glx) depending on your OS
• MCP client (e.g., Claude, Cline, or Cursor) to connect to the MCP server

Installation steps

1. Ensure prerequisites are installed

• On Debian/Ubuntu:

sudo apt-get update
sudo apt-get install -y python3 python3-venv ffmpeg libsm6 libxext6 libgl1-mesa-glx

• Install pipx if not present:

python3 -m pip install --user pipx
python3 -m pipx ensurepath

2. Install ImageSorcery MCP via pipx

pipx install imagesorcery-mcp

3. Run the MCP server locally

# Run the server (this uses pipx to execute the package entrypoint)
pipx run imagesorcery-mcp

4. Add the server to your MCP configuration

• In your MCP configuration file, add the server under mcpServers:

"mcpServers": {
  "imagesorcery-mcp": {"command": "pipx", "args": ["run", "imagesorcery-mcp"]}
}

5. Optional: enable telemetry or adjust settings

• Use the config tool to modify settings (e.g., set telemetry.enabled to true/false) as described in the README of the server.

Notes

• If you use Docker or a virtual environment, ensure system libraries required by OpenCV are accessible in the container or environment.
• The server exposes a rich set of tools (crop, resize, detect, ocr, draw_* tools, etc.). See the Imagesorcery MCP docs in the repository for details on each tool’s usage.

Tips and common issues:

• Ensure Python 3.10+ and pipx are installed; if you encounter import or binary issues, reinstall the MCP package with pipx.
• OpenCV-based tools require system libraries; on headless servers, ensure fonts and display-related dependencies are handled or avoided.
• When using OCR or model-based tools, provide full paths to images to avoid path resolution issues.
• Telemetry: the server can be configured to send usage telemetry; disable it if running in air-gapped environments. See the config tool in the MCP UI for details.
• If your MCP client cannot reach the local server, verify that the MCP configuration file is loaded correctly and that the server process is running.