ImageMcp

A full-featured image processing MCP server for AI assistants. Exposes ~55 tools across 11 categories — editing, layers, format conversion, AI segmentation/cleanup/generation, design analysis, screenshot-to-code, and more.

Quick Start

# Install
pip install -e .

# Set API key (required for AI-powered tools)
export ANTHROPIC_API_KEY="sk-..."

# Run the MCP server
python server.py

# Or with the MCP CLI
mcp run server.py

Without ANTHROPIC_API_KEY, all non-AI tools work (core editing, layers, conversions) and AI tools degrade to local Pillow fallbacks with reduced quality.

Tools

Core Editing (10)

crop_image, resize_image, rotate_image, flip_image, add_text, remove_text, blur_region, adjust_brightness, adjust_contrast, export_image

Layer Management (8)

create_document, add_image_layer, add_text_layer, move_layer, resize_layer, delete_layer, duplicate_layer, list_layers

Format Conversions (7)

png_to_jpg, jpg_to_png, webp_to_png, svg_to_png, png_to_svg, image_to_pdf, pdf_to_images

AI Segmentation & Selection (6)

extract_subject, extract_person, extract_face, extract_object, remove_background, generate_mask

AI Cleanup (4)

remove_object, erase_text, remove_watermark_candidate, inpaint_region

AI Generation (5)

generate_avatar, generate_icon, generate_background, generate_illustration, generate_character

Design Analysis (5)

extract_colors, extract_typography, detect_layout, describe_design, identify_components

Screenshot → Code (4)

screenshot_to_html, screenshot_to_react, screenshot_to_component_tree, image_to_wireframe

Smart Export (5)

export_png, export_svg, export_react, export_tailwind, export_figma_json

Advanced AI (7)

photo_to_headshot, photo_to_cartoon, photo_to_vector, photo_to_3d, style_transfer, face_enhancement, upscale_image

Architecture

D:\ImageMcp\
├── server.py                    # FastMCP entry — registers all 55 tools
├── main.py                      # CLI entry point
├── pyproject.toml               # Python project config + dependencies
│
├── src/imagemcp/
│   ├── tools/                   # One module per tool category
│   │   ├── core_editing.py
│   │   ├── layers.py
│   │   ├── conversions.py
│   │   ├── ai_segmentation.py
│   │   ├── ai_cleanup.py
│   │   ├── ai_generation.py
│   │   ├── design_analysis.py
│   │   ├── screenshot_to_code.py
│   │   ├── smart_export.py
│   │   └── advanced_ai.py
│   │
│   └── utils/
│       ├── io.py                # Image I/O, temp file management
│       ├── ai_client.py         # Anthropic SDK client, vision helpers, image generation
│       └── canvas.py            # In-memory layer canvas for compositing
│
└── tests/                       # ~120 tests across all tool categories
    ├── conftest.py
    └── test_*.py

Configuration

Connecting to the Server

Once the server is running, any MCP-compatible client can connect via stdio transport.

Claude Desktop / Claude Code

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "ImageMcp": {
      "command": "python",
      "args": ["D:/ImageMCP/server.py"]
    }
  }
}

VS Code (GitHub Copilot)

Create or edit .vscode/mcp.json in your workspace:

{
  "servers": {
    "ImageMcp": {
      "type": "stdio",
      "command": "python",
      "args": ["D:/ImageMCP/server.py"]
    }
  }
}

Using UV

If you use uv to manage the project:

{
  "mcpServers": {
    "ImageMcp": {
      "command": "uv",
      "args": ["run", "server.py"],
      "cwd": "D:/ImageMCP"
    }
  }
}

Custom MCP Client (stdio)

The server communicates over stdin/stdout using the Model Context Protocol (MCP) JSON-RPC format. Any MCP-compatible client can connect — no HTTP server needed.

Development

# Install dev dependencies
pip install -e ".[test]"

# Download test assets
python -m tests.download_assets

# Run tests (API tests auto-skip if ANTHROPIC_API_KEY not set)
pytest tests/ -v

Stack

• MCP framework: mcp[cli] (FastMCP)

• Image processing: Pillow, numpy

• AI: Anthropic Claude SDK (vision, image generation, inpainting)

• Background removal: rembg (U²-Net, runs locally)

• Format support: cairosvg, PyMuPDF, reportlab

• OCR: pytesseract (optional)