pixibox-mcp-server
v2.2.2
Published
Model Context Protocol server for Pixibox.ai — generate 3D models, control DCC tools (Blender/Houdini/Maya/UE5), manage assets, and orchestrate production pipelines via AI agents.
Maintainers
Readme
Pixibox MCP Server v2.0
Model Context Protocol server for Pixibox.ai — generate 3D models, control DCC tools, manage assets, and orchestrate production pipelines via AI agents.
55+ Tools
| Category | Tools | Description | |----------|-------|-------------| | Generation | 8 | Create 3D from image/text, poll status, export formats | | Batch | 6 | Batch generation, retry, cancel | | DCC Bridge | 5 | Control Blender/Houdini/Maya/UE5 — push models, execute scripts | | DCC Helpers | 3 | Pre-built: Blender render, scene info, Houdini HDA cook | | Pipeline | 6 | Multi-step workflows: image→3D→optimize→DCC | | Assets | 4 | Upload, download, list, delete cloud assets | | Textures | 3 | AI texture generation, PBR presets, style transfer | | Optimization | 1 | Mesh decimation & LOD | | Rigging | 5 | Auto-rig, pose, animate | | Arena | 4 | Leaderboard, battles, voting | | Gallery | 4 | Browse, publish, tag | | Teams | 4 | Create teams, share assets | | Collaboration | 6 | Comments, annotations, reviews, versioning | | Account | 7 | Usage, subscription, notifications, AI models list | | Image Gen | 1 | Hunyuan image generation |
Quick Start
With Claude Code
npm install -g pixibox-mcp-serverAdd to ~/.claude/claude_desktop_config.json:
{
"mcpServers": {
"pixibox": {
"command": "npx",
"args": ["-y", "-p", "pixibox-mcp-server", "pixi-mcp-server"],
"env": {
"PIXIBOX_API_KEY": "px_your_key_here"
}
}
}
}With Claude Desktop (SSE)
MCP_TRANSPORT=sse MCP_PORT=3100 PIXIBOX_API_KEY=px_... \
npx -y -p pixibox-mcp-server pixi-mcp-serverPixibox CLI
export PIXIBOX_API_KEY=px_your_key_here
# Generate 3D from image
pixibox generate --image https://example.com/photo.jpg --model trellis --wait
# Check DCC bridge
pixibox bridge status
# Push model to Blender
pixibox bridge push blender --generation gen_abc123
# Run pipeline
pixibox pipeline run image-to-dcc --image https://example.com/photo.jpg
# Start MCP server
pixibox mcp
pixibox mcp --sse --port 3100Environment Variables
| Variable | Required | Default | Description |
|----------|----------|---------|-------------|
| PIXIBOX_API_KEY | Yes | — | API key (px_...) |
| PIXIBOX_API_BASE | No | https://pixibox.ai/api | API base URL |
| MCP_TRANSPORT | No | stdio | stdio or sse |
| MCP_PORT | No | 3100 | SSE server port |
Pipeline Templates
| ID | Steps |
|----|-------|
| image-to-dcc | Generate → Push to DCC |
| text-to-optimized | Generate → Optimize mesh |
| full-pipeline | Generate → Optimize → Export → Push to DCC |
| batch-compare | Generate (Trellis + Hunyuan + Rodin) |
| houdini-procedural | Generate → Push to Houdini → Run HDA |
Architecture
Claude Code / Claude Desktop / Pixibox CLI
↓ (stdio or SSE)
Pixibox MCP Server
↓ (REST API)
Pixibox Backend (Express + Socket.IO)
├→ BullMQ → AI Providers (Fal.ai, NVIDIA, Hunyuan)
├→ GCS Storage
└→ Socket.IO /bridge → Blender / Houdini / Maya / UE5License
MIT — Aerovfx / Pixibox.ai
