Open Source · Model Context Protocol · MIT

Batch-process a whole folder of images from inside an AI agent, privately and offline

image-toolkit-mcp is a local MCP server built on Pillow that resizes, converts, compresses, watermarks and strips EXIF from individual files or entire folders — 100% on the machine, no API key, images never uploaded.

Why an MCP server for this

Online image processors require uploading every photo, which is unacceptable for client, medical or personal images, and manual editors do not scale to whole folders. This server gives an agent a callable local tool that processes an entire directory in one request with predictable, scriptable results.

E-commerce and marketing

Resize and compress a folder of product photos to platform specs and add a brand watermark in one batch, without uploading them.

Web performance

Convert assets to WebP and tune quality to hit size budgets across an entire image directory.

Privacy and compliance

Strip EXIF GPS, device and timestamp metadata before sharing or archiving client and personal photos.

AI agents and automation

Give Claude Desktop or Cursor a local image tool for repeatable resize/convert pipelines with no third-party service.

What it does

Install and connect

Published to the official MCP Registry as io.github.diyaluo/image-toolkit-mcp and as an OCI image on GHCR. Run locally over stdio:

pip install -e .
image_toolkit_mcp

Client configuration (Claude Desktop / Cursor / any MCP client):

{
  "mcpServers": {
    "image-toolkit": {
      "command": "image_toolkit_mcp",
      "env": {}
    }
  }
}

No API key and no network. Install Pillow and run over stdio; images are read from and written to local directories only and never uploaded.

Frequently asked questions

What does this MCP server do?

It exposes batch image processing as MCP tools, so an MCP-compatible agent can resize, convert, compress, watermark and strip EXIF from a single image or an entire local folder.

Are my images uploaded anywhere?

No. The server runs locally on Pillow and communicates over stdio. Images never leave the machine, which makes it suitable for confidential client and personal photos.

Which image formats are supported?

Conversion and processing cover JPG/JPEG, PNG, WebP, BMP and TIFF, with sensible handling of RGBA and palette-mode images when saving to JPEG.

What does EXIF stripping remove?

image_strip rewrites each image without its EXIF block, removing camera model, GPS location and capture timestamp so shared files do not leak metadata.

Which MCP clients are supported?

Any Model Context Protocol client over stdio, including Claude Desktop and Cursor. It is published to the official MCP Registry as io.github.diyaluo/image-toolkit-mcp and on GHCR.