ChangeImageTo MCP — Image editing for Cursor, Claude & any MCP client

**A production-ready Model Context Protocol (MCP) server for AI-assisted image editing.** Give Cursor, Claude Desktop, Windsurf, VS Code MCP, and other MCP hosts 16 tools to remove backgrounds, blur or grayscale backgrounds, change colors, convert formats (PNG/JPEG/WebP/…), upscale, denoise/enhance, run OCR, export PDF, edit text in images (Gemini), clean watermarks, inpaint masked regions, and run bulk resize / convert / quality checks.

> Search-friendly aliases: image editing MCP · MCP image tools · background removal MCP · remove background API MCP · photo editing LLM tools · Cursor image MCP · Claude image MCP · batch image processing MCP · image upscaling MCP · OCR MCP server.

The server is a small HTTP client: it forwards requests to the same ChangeImageTo image API used by changeimageto.com. Point it at the hosted backend (default) or at your own deployment of the matching `/api/*` routes.

Why this repo?

GitHub topics (repository tags)

GitHub allows 20 topics per repository. This repo’s topics (for search and discoverability):

`mcp` · `model-context-protocol` · `mcp-server` · `image-editing` · `image-processing` · `background-removal` · `remove-background` · `photo-editing` · `image-tools` · `cursor` · `claude` · `claude-desktop` · `llm` · `ai-tools` · `python` · `ocr` · `image-upscaling` · `inpainting` · `fastapi` · `changeimageto`

Install

git clone https://github.com/vipul510-web/mcp-changeimageto.git
cd mcp-changeimageto
python3 -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install .

Install from Git without cloning:

pip install git+https://github.com/vipul510-web/mcp-changeimageto.git

Cursor, Claude Desktop, VS Code — MCP config

Replace paths with yours:

{
  "mcpServers": {
    "changeimageto": {
      "command": "/full/path/to/mcp-changeimageto/.venv/bin/python",
      "args": ["/full/path/to/mcp-changeimageto/server.py"],
      "env": {
        "CHANGEIMAGETO_API_BASE": "https://bgremover-backend-121350814881.us-central1.run.app"
      }
    }
  }
}

Or use the console script after `pip install .`:

{
  "mcpServers": {
    "changeimageto": {
      "command": "/full/path/to/.venv/bin/mcp-changeimageto"
    }
  }
}

The entrypoint runs `mcp.run()` (stdio transport).

Environment variables

Image inputs

Each tool accepts `image` (or a list for bulk tools) as:

• HTTPS URL to an image
• Local file path
• Data URL: `data:image/png;base64,...`

Responses that are binary (PNG, JPEG, PDF, ZIP) are returned as JSON with `data_base64` when the payload is under ~12 MB. Larger outputs omit base64 and include a short notice—use a local API, smaller files, or non-bulk endpoints.

Tools ↔ REST API

Test-only, blog, analytics, and payment endpoints are not wrapped (by design).

Related

• Product & docs: changeimageto.com
• MCP spec: modelcontextprotocol.io
• Issues & ideas: github.com/vipul510-web/mcp-changeimageto/issues

License

MIT