Pixibox AI MCP Server

A Model Context Protocol (MCP) server for Pixibox.ai — the 3D AI comparison platform. Generate 3D models from images or text using multiple AI providers, compare results in the arena, collaborate with teams, and manage your assets — all from MCP clients like Claude Desktop, Cursor, or VS Code.

Features

• Image-to-3D — Generate 3D models from reference images
• Text-to-3D — Create 3D models from text prompts
• Multi-model comparison — Use Trellis, Hunyuan3D, Rodin, TripoSG, and more
• Batch generation — Submit multiple jobs at once
• Texture generation — AI textures, style transfer, PBR presets
• Model optimization — Mesh decimation, LOD generation
• Rigging & Animation — Auto-rig characters and apply animations
• Arena & Leaderboard — Compare AI models via ELO-rated battles
• Gallery — Browse and publish public 3D assets
• Teams & Collaboration — Share assets, add comments, 3D annotations, review workflows
• Versioning — Track iterations of your generations
• Polling — Built-in polling for async generation and batch tasks

Installation

Option 1: Run with npx (Recommended)

{
  "mcpServers": {
    "pixibox-ai": {
      "command": "npx",
      "args": ["-y", "pixi-mcp-server"],
      "env": {
        "PIXIBOX_API_KEY": "your_api_key_here"
      }
    }
  }
}

Option 2: Clone and Build Locally

git clone https://github.com/aerovfx/pixi-mcp-server.git
cd pixi-mcp-server
npm install
npm run build

Then configure your MCP client:

{
  "mcpServers": {
    "pixibox-ai": {
      "command": "node",
      "args": ["/absolute/path/to/pixi-mcp-server/dist/index.js"],
      "env": {
        "PIXIBOX_API_KEY": "your_api_key_here"
      }
    }
  }
}

Configuration

Get your API key from Pixibox.ai Settings → API Keys.

Environment Variables

Available Tools

Generation (Core)

Batch

Textures

Model Optimization

Rigging & Animation

Arena & Leaderboard

Gallery

Scenes

Teams & Collaboration

Versioning

Account & Utility

Image Generation

Example Usage

Generate a 3D model from text

Use create_generation with model "trellis" and prompt "a cute robot character"
Then use wait_for_generation with the returned generation_id

Compare two AI models

Use create_batch with items:
  - model: "trellis", prompt: "medieval castle"
  - model: "hunyuan3d-2", prompt: "medieval castle"
Then use wait_for_batch to get both results

Browse the leaderboard

Use get_leaderboard with category "image_to_3d" to see which AI model is best

Development

echo "PIXIBOX_API_KEY=your_key_here" > .env
npm run dev

License

MIT