@topazlabs/mcp

AI image enhancement for LLMs. An MCP server that lets Claude and other AI assistants enhance images using Topaz Labs AI models.

Topaz Labs makes AI-powered image and video enhancement software used by photographers, filmmakers, and developers worldwide.

See It in Action

Before	After

Enhanced with the Wonder 2 model. View interactive comparison →

Quick Start

1. Get an API key

2. Add to your MCP client

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):

hljs language-json

{
  "mcpServers": {
    "Topaz Labs": {
      "command": "npx",
      "args": ["-y", "@topazlabs/mcp"],
      "env": {
        "TOPAZ_API_KEY": "<your-api-key>"
      }
    }
  }
}

3. Enhance images

You: "Enhance ~/Desktop/photo.jpg with Wonder 2"

Claude: I'll enhance your photo using Topaz Labs' Wonder 2 model.

[calls image-enhance tool]

Note: This tool requires a file path or URL — it cannot process images uploaded directly in the chat. It works best with Claude Cowork, which has file system access. It does not work in Claude Chat mode (no file system access).

Here's your enhanced image -- upscaled and sharpened with improved detail and clarity.

Compatibility

This server uses stdio transport, supported by all major MCP desktop clients:

Client	Status
Claude Cowork	✅ (recommended — has file system access)
Claude Desktop (Chat)	⚠️ Requires user to provide file paths manually
Cursor	✅
Windsurf	✅
VS Code (Copilot)	✅
Cline	✅
Claude Code	✅
Zed	✅

HTTP streaming transport (for ChatGPT, remote agents, web/mobile clients) is not yet supported but may be added in a future release.

Tool

`image-enhance`

Send an image, get back an enhanced version. Defaults to Standard V2 and auto-upscales to 2K.

hljs language-vbnet

image: "~/Photos/photo.jpg"     # Local path or URL (required)
size: { scale: 2 }              # Optional: exact dims, single dim, or scale factor
model: "Standard V2"            # Optional: defaults to Standard V2

Size options: { width, height } for exact, { width } or { height } for proportional, { scale } for a multiplier. If omitted, images under 2K are auto-upscaled to 2K width.

Models:

Model	Type	Best for
Standard V2	standard	General photos (default)
Low Resolution V2	standard	Tiny images, thumbnails
CGI	standard	Digital art, renders, illustrations
High Fidelity V2	standard	High-quality sources needing upscale
Redefine	generative	Creative reinterpretation
Recover 3	generative	Low-res images, old photos
Standard MAX	generative	Maximum quality, slower
Wonder 2	generative	Nature photography, portraits, fine detail

Environment Variables

Variable	Required	Description
`TOPAZ_API_KEY`	Yes	Topaz Labs API key

Architecture

hljs language-lua

src/
  api/        -- Topaz API client, types, model definitions
  lib/        -- Path security, URL fetching utilities
  tools/      -- enhance tool handler
  server.ts   -- MCP server setup and tool registration
  main.ts     -- Entry point (stdio transport)
test/         -- Vitest test suite

Development

hljs language-bash

npm install
npm run build
npm test

Limitations

Image only -- video enhancement is not yet supported
Paid API key required -- sign up at developer.topazlabs.com
Cloud processing -- images are sent to the Topaz Labs API for enhancement

Contributing

Found a bug or have a feature request? Open an issue on GitHub.

License

MIT -- Topaz Labs

topaz-mcp