CLO3D MCP Server
Enables AI assistants to control CLO3D for pattern creation, fabric swapping, simulation, and model export via natural language commands.
README
CLO3D MCP Server
Talk to CLO3D from Claude, Cursor, or any MCP client. Create patterns, swap fabrics, run simulations, export models. All from a chat window.
If you use CLO3D and you've wished you could script repetitive tasks without digging through the API docs, this is for you.
Your AI assistant <--> MCP Server <--> CLO3D Plugin <--> CLO3D
Quick Start
1. Load the plugin in CLO3D
One Python file, no dependencies.
- Open CLO3D
- Script > Script Editor
- Open
plugin/clo3d_mcp_plugin.py - Hit Run
You'll see [CLO MCP] Plugin started in the console.
2. Connect your AI client
Claude Code:
claude mcp add clo3d -- uvx clo3d-mcp
Claude Desktop (add to claude_desktop_config.json):
{
"mcpServers": {
"clo3d": {
"command": "uvx",
"args": ["clo3d-mcp"]
}
}
}
3. Go
Ask things like:
- "What patterns are in this project?"
- "Set the front bodice fabric to red"
- "Run 100 steps of simulation then export as GLB"
- "Export a turntable of all colorways"
What you get (33 tools)
Scene
| Tool | What it does |
|---|---|
get_project_info |
Project name, file path, CLO version, piece counts |
new_project |
Start fresh |
open_file |
Open .zprj, .zpac, .obj, .fbx, .avt |
save_project |
Save as .zprj |
get_garment_info |
Dump full garment metadata to JSON |
Patterns
| Tool | What it does |
|---|---|
get_pattern_count |
How many pieces |
get_pattern_list |
All pieces with names and indices |
get_pattern_info |
Full detail on one piece |
get_pattern_bounding_box |
Width and height |
set_pattern_name |
Rename a piece |
copy_pattern |
Duplicate at a given position |
delete_pattern |
Remove a piece |
flip_pattern |
Flip horizontal or vertical |
create_pattern |
Build a new piece from vertex points |
get_arrangement_list |
Get avatar arrangement points |
Fabrics
| Tool | What it does |
|---|---|
get_fabric_list |
All fabrics with indices |
add_fabric |
Load a .zfab or .jfab |
replace_fabric |
Swap an existing fabric |
assign_fabric_to_pattern |
Put a fabric on a pattern piece |
set_fabric_color |
Set PBR base color (RGB) |
get_fabric_for_pattern |
Check which fabric is on a piece |
Export
| Tool | What it does |
|---|---|
export_obj |
OBJ with options |
export_fbx |
FBX |
export_glb |
GLB (binary glTF) |
export_gltf |
glTF |
export_thumbnail |
3D viewport screenshot |
export_snapshot |
Multi-view snapshots |
export_turntable |
360 turntable image sequence |
export_tech_pack |
Tech pack with JSON + images |
Import
| Tool | What it does |
|---|---|
import_file |
Auto-detect and import any supported format |
Simulation
| Tool | What it does |
|---|---|
simulate |
Run cloth sim for N steps |
Colorways
| Tool | What it does |
|---|---|
get_colorways |
List all colorways |
set_current_colorway |
Switch the active colorway |
How it works
Two pieces:
The plugin (plugin/clo3d_mcp_plugin.py) runs inside CLO3D. It's a single Python file that polls a shared temp directory for JSON commands, calls the CLO3D Python API, and writes results back. No external dependencies, no sockets, works with CLO3D's built-in Python.
The MCP server (src/clo3d_mcp/) is a Python package that writes commands to that directory and reads responses. Any MCP-compatible client (Claude, Cursor, etc.) can use it.
┌─────────────────┐ file-based IPC ┌──────────────────┐ CLO3D API ┌────────┐
│ MCP Server │ ◄──────────────► │ CLO3D Plugin │ ◄──────────────► │ CLO3D │
│ (clo3d-mcp) │ %TEMP%/clo3d_mcp │ (Python script) │ │ │
└─────────────────┘ └──────────────────┘ └────────┘
Troubleshooting
"Cannot find CLO3D communication directory": Is CLO3D running? Did you run the plugin script?
Plugin errors: Check the Script Editor console in CLO3D for Python tracebacks.
Simulation/export timeout: These can take a while. The default timeout is 180 seconds.
Custom comm directory: Set the CLO3D_MCP_DIR environment variable to override the default shared directory path.
Requirements
- CLO3D with Python scripting support
- Python 3.10+
- uv or pip
Development
git clone https://github.com/Ubani-Studio/clo3d-mcp.git
cd clo3d-mcp
# Install dev dependencies
uv sync --dev
# Run tests (no CLO3D needed, uses a mock server)
uv run pytest tests/ -v
# Run the MCP server locally
uv run clo3d-mcp
License
MIT. See LICENSE.
CLO3D SDK headers in sdk/ are provided by CLO Virtual Fashion for plugin development. You need your own CLO3D license.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。