Version: 1.0.0 | Date: 16.02.2026 | Time: 12:00 | GlobalID: 20260216_1200_USDcodeNIM_MCP_batch

Tag block: #framework_integration #stage #v_model #aas_integration #cursor #vscode #validation #specializes #mcp_protocol #usd_core #opencode #openusd #semantic_governance #export #workflow_automation #ai_coding_agents #quality_assurance #deterministic_workflows #best_practices

NVIDIA NIM (NVIDIA Intelligent Model) integration for Cursor IDE, specifically configured for USD (Universal Scene Description) code generation and validation.

This repository provides tools and configurations to integrate NVIDIA's specialized USD code model into Cursor IDE through the Model Context Protocol (MCP). It enables:

• USD Code Generation using NVIDIA's specialized USD code model
• Automated Code Validation with structured feedback
• USD-Specific Suggestions and best practices
• Double-Checking generated code for errors and improvements

1. Get your API key

   1. from build.nvidia.com/nvidia/usdcode

   Navigate to the NVIDIA NIM platform and search for "USD code" or "usdcode" in the model catalog. Once you find the USD code model, sign in or create an account to access your API key. The API key will be displayed on the model's page after authentication.

   2. Set environment variable:

   $env:NIM_API_KEY = "your_api_key_here"  # Windows PowerShell

   set NIM_API_KEY=your_api_key_here  # Windows CMD

   export NIM_API_KEY=your_api_key_here  # Linux/Mac

   3. Install dependencies:

   pip install -r scripts/requirements_nim.txt

   4. Test the integration:

   python scripts/nim_mcp_server.py test

Open Cursor and click the cog wheel icon in the upper right corner to access settings. Select Cursor Settings, then navigate to Tools & MCP. Click Add a new MCP server and paste the following JSON configuration into the file:

{
  "mcpServers": {
    "nvidia-nim": {
      "command": "python",
      "args": ["C:\\path\\to\\USDcodeNIM_MCP\\scripts\\nim_mcp_server.py"],
      "env": {
        "NIM_API_KEY": "your_api_key_here",
        "NIM_MODEL": "nvidia/usdcode-llama-3.1-70b-instruct"
      }
    }
  }
}

Make sure to update the path to match your repository location and replace your_api_key_here with your actual API key.

1. Restart Cursor Close and reopen Cursor completely.

2. Verify Connection Settings → Tools & MCP → Check nvidia-nim is green ✅

💡 Documentation Hierarchy:

• This README → Quick Start (basic setup)
• Quick Reference Guide → More depth, troubleshooting, and examples
• Complete Guide → Everything described in detail

USDcodeNIM_MCP/
├── scripts/
│   ├── nim_mcp_server.py      # MCP server wrapper for Cursor integration
│   ├── nim_direct_client.py    # Direct client for standalone use
│   ├── validate_with_nim.py   # Validation script for USD Python files
│   ├── requirements_nim.txt    # Python dependencies
│   └── README_NIM.md          # Quick reference guide
├── config/
│   └── cursor_mcp_config.json  # Cursor MCP configuration template
├── docs/
│   └── NVIDIA_NIM_Integration_Guide.md  # Complete setup documentation
└── README.md                  # This file

Full Cursor IDE integration through Model Context Protocol, providing USD code generation and validation tools.

Standalone Python client for use in scripts, CI/CD pipelines, or other tools.

Automated validation of USD Python code with structured feedback (errors, warnings, suggestions).

Generate USD code from natural language prompts using NVIDIA's specialized model.

Once configured, use NVIDIA NIM directly in Cursor's chat:

Generate USD Code:

Generate a USD Python script that creates a red cube mesh

Validate USD Code:

Validate this USD script using NVIDIA NIM

(With a USD file open)

Ask USD Questions:

What's the best way to create a USD stage?

python scripts/validate_with_nim.py path/to/usd_script.py
python scripts/validate_with_nim.py path/to/usd_script.py --context "Asset validation script"

python scripts/nim_direct_client.py "Create a USD stage with a cube mesh"

from scripts.nim_direct_client import NIMDirectClient
import asyncio

async def main():
    client = NIMDirectClient()
    
    # Generate code
    code = await client.generate_code("Create a USD stage")
    print(code)
    
    # Validate code
    result = await client.validate_code(code)
    print(result)
    
    await client.close()

asyncio.run(main())

To integrate with Cursor IDE:

1. Get your API key from build.nvidia.com/nvidia/usdcode
2. Edit your MCP config: C:\Users\<username>\.cursor\mcp.json
3. Add the NVIDIA NIM server configuration from config/cursor_mcp_config.json
4. Update the path in the config to match your repository location
5. Set your API key in the env.NIM_API_KEY field
6. Set model name: nvidia/usdcode-llama-3.1-70b-instruct
7. Restart Cursor completely

Quick Setup:

{
  "mcpServers": {
    "nvidia-nim": {
      "command": "python",
      "args": ["C:\\path\\to\\USDcodeNIM_MCP\\scripts\\nim_mcp_server.py"],
      "env": {
        "NIM_API_KEY": "your_api_key_here",
        "NIM_MODEL": "nvidia/usdcode-llama-3.1-70b-instruct"
      }
    }
  }
}

Need more help?

• For more depth and troubleshooting → See QUICK_REFERENCE.md
• For comprehensive documentation → See docs/NVIDIA_NIM_Integration_Guide.md

• Python 3.10+
• httpx library (install via pip install -r scripts/requirements_nim.txt)
• NVIDIA NIM API key from build.nvidia.com
• Cursor IDE (for MCP integration)

Documentation Hierarchy:

1. README.md (you are here) → Quick Start with basic setup instructions
2. QUICK_REFERENCE.md → More depth, troubleshooting, examples, and common commands
3. Complete Guide → Comprehensive documentation with everything described in detail

Additional Resources:

• Scripts Guide: scripts/README_NIM.md - Script-specific documentation
• Configuration: config/cursor_mcp_config.json - MCP configuration template

• NVIDIA NIM USD Code Model
• NVIDIA NIM Documentation
• MCP Protocol Specification
• USD Python API Documentation

This project is provided as-is for integration with NVIDIA NIM services.

Version: 1.0.0
Last Updated: 2025-11-21