MCP Face Transform Server

A Model Context Protocol (MCP) server that exposes the AceDataCloud Face Transform API — face keypoint detection, beautification, age/gender transform, face swap, cartoonization, and liveness detection.

Status: All Face APIs are currently in Alpha. Interfaces may evolve.

Features

• Keypoint detection — 90+ landmarks per face, multi-face supported
• Beautification — smoothing, whitening, face slimming, eye enlarging
• Age transform — age or de-age a portrait
• Gender transform — swap perceived facial gender characteristics
• Face swap — move a source face onto a target image (with optional async webhook)
• Cartoonize — render a portrait in animated / cartoon style
• Liveness detection — distinguish live captures from printed / screen photos

Installation

pip install mcp-face-transform

Configuration

Set your AceDataCloud API token:

export ACEDATACLOUD_API_TOKEN=your_token_here

Get your token from https://platform.acedata.cloud.

Usage

stdio mode (default)

mcp-face-transform

HTTP mode

mcp-face-transform --transport http --port 8000

Tool Reference

Example

"Detect all faces in https://example.com/group.jpg and return their keypoints."
→ face_detect_keypoints(image_url="https://example.com/group.jpg")

"Lighten and smooth my portrait."
→ face_beautify(image_url="https://example.com/me.jpg", smoothing=15, whitening=25)

"Replace the face in the scene with the headshot."
→ face_swap(
    source_image_url="https://example.com/headshot.jpg",
    target_image_url="https://example.com/scene.jpg",
  )

Configuration in Claude Desktop / Claude Code

{
  "mcpServers": {
    "face-transform": {
      "command": "uvx",
      "args": ["mcp-face-transform"],
      "env": {
        "ACEDATACLOUD_API_TOKEN": "your_api_token_here"
      }
    }
  }
}

Or use the hosted endpoint with bearer auth:

{
  "mcpServers": {
    "face-transform": {
      "url": "https://face.mcp.acedata.cloud/mcp",
      "headers": {
        "Authorization": "Bearer your_api_token_here"
      }
    }
  }
}

Development

pip install -e ".[dev,test]"
pytest --cov=core --cov=tools
ruff check .

License

MIT — see LICENSE.