Claude Code for RTX Tool Version Manager Workflow

Managing GPU tool versions across different projects can be a headache. Whether you’re switching between CUDA versions, matching TensorFlow/PyTorch builds, or maintaining compatibility with specific driver versions, the RTX Tool Version Manager workflow powered by Claude Code can transform this tedious process into an automated, error-free experience.

Understanding RTX Tool Version Manager

The RTX Tool Version Manager is a CLI utility designed specifically for developers working with NVIDIA RTX GPUs. It handles version switching, dependency resolution, and environment configuration for CUDA toolkit, cuDNN, and related libraries. When integrated with Claude Code, it becomes part of an intelligent workflow that understands your project requirements and automatically configures the right tool versions.

The core philosophy is simple: specify what you need, and let Claude Code and RTX Tool Version Manager handle the rest. No more manual environment toggling or compatibility hunting.

Setting Up the Integration

Before diving into workflows, ensure you have the RTX Tool Version Manager installed and accessible. Here’s how to verify your setup:

# Check RTX Tool Version Manager installation
rtx --version

# List available CUDA versions
rtx list cuda

# Display current active versions
rtx current

Once verified, you can invoke Claude Code with specific context about your tool version needs. The key is providing Claude with clear information about your project requirements.

Workflow 1: Project-Based Version Switching

The most common use case is switching tool versions when moving between projects. Each project typically has specific version requirements documented in a .rtx-version file or similar configuration.

Here’s a practical workflow:

# In your project directory, create a version file
echo "cuda=11.8.0" > .rtx-version
echo "cudnn=8.9.0" >> .rtx-version
echo "tensorrt=8.6.0" >> .rtx-version

When you ask Claude Code to help with this project, it can read these requirements and automatically configure your environment:

“Help me set up the development environment for this project. I need to ensure CUDA 11.8 and cuDNN 8.9 are properly configured.”

Claude Code will read your .rtx-version file, invoke rtx install for each tool, and verify the installation by checking nvcc --version and library paths.

Workflow 2: Automatic Dependency Resolution

One of RTX Tool Version Manager’s powerful features is automatic dependency resolution. When you request a specific CUDA version, it can automatically install compatible cuDNN and TensorRT versions. Claude Code enhances this by understanding your broader context.

Consider this scenario: you need to set up a PyTorch environment with GPU support. Instead of manually researching version compatibility, you can ask Claude:

“Set up PyTorch 2.1 with CUDA 12.1 support. Ensure all dependencies are compatible.”

Claude Code will:

Query available CUDA versions via RTX
Install CUDA 12.1 if not present
Install cuDNN version compatible with CUDA 12.1
Install PyTorch with the correct CUDA bindings
Verify the setup with a quick test script

import torch
print(f"PyTorch version: {torch.__version__}")
print(f"CUDA available: {torch.cuda.is_available()}")
print(f"CUDA version: {torch.version.cuda}")

Workflow 3: Environment Validation and Debugging

When things don’t work as expected, Claude Code can help diagnose version conflicts. The workflow involves checking current versions, comparing against requirements, and identifying mismatches.

# Claude Code can run these diagnostic commands
rtx list --installed
rtx list cuda --all
ldconfig -p | grep cuda

If you’re experiencing issues, provide Claude with the error output. It can then cross-reference your installed versions with requirements and suggest specific version changes.

Common issues include:

Driver compatibility: CUDA version exceeds driver support
Symbol conflicts: Multiple CUDA versions in library path
Missing dependencies: cuDNN or TensorRT not properly linked

Workflow 4: Multi-Project Environment Management

Developers often work on multiple projects simultaneously, each with different version requirements. The RTX Tool Version Manager supports .rtx-version files per directory, enabling seamless switching.

Claude Code can manage this workflow intelligently:

# Activate project-specific versions
cd ~/projects/ml-model-a
rtx activate

# Claude Code remembers context per project
# When you switch projects, it recognizes the new .rtx-version
cd ~/projects/inference-service-b
rtx activate

You can also create aliases for frequently used combinations:

# Create a preset for common setups
rtx alias set ml-training "cuda=12.1 cudnn=8.9 tensorrt=8.5"
rtx alias set inference "cuda=11.8 cudnn=8.6 tensorrt=8.4"

Best Practices for Claude Code Integration

To get the most out of this workflow, follow these best practices:

1. Document Version Requirements Clearly

Always maintain a .rtx-version file in your project root:

# .rtx-version for my-project
cuda=12.1.0
cudnn=8.9.0
tensorrt=8.5.1

2. Use Semantic Versioning

When specifying versions, use exact versions for reproducibility:

# Good: exact version
rtx install cuda=12.1.0

# Avoid: moving target
rtx install cuda@latest

3. Verify Before Building

Always confirm the environment is correctly set up before starting builds:

# Quick verification script
nvcc --version
echo $CUDA_HOME
ldconfig -p | grep cudnn

4. Leverage Claude’s Context

When asking for help, provide context:

Current installed versions (rtx current)
Your target version from project requirements
Any error messages you’re seeing

Advanced: Custom Skills for Version Management

You can create a Claude Code skill that encapsulates your version management workflow. A custom skill can automate common tasks:

---
name: rtx-manager
description: Manage RTX tool versions and GPU environments
---

# RTX Version Manager Helper

This skill assists with GPU tool version management using RTX.

## Available Commands

- `rtx list <tool>` - List available versions
- `rtx install <tool>=<version>` - Install specific version
- `rtx activate` - Activate versions from .rtx-version
- `rtx current` - Show currently active versions

## Common Tasks

### Install CUDA Version
Run: `rtx install cuda=<version>`

### Switch to Project Versions
Run: `rtx activate` in project directory

### Verify Installation
Run verification commands to confirm proper setup.

Conclusion

The RTX Tool Version Manager workflow combined with Claude Code creates a powerful automation system for GPU development environments. By documenting your requirements, using automatic dependency resolution, and providing clear context to Claude, you can eliminate manual version management headaches and focus on what matters: building great software.

Remember to keep your .rtx-version files updated, verify environments before major builds, and let Claude Code handle the complexity of version compatibility. Your future self will thank you when switching between projects becomes as simple as changing directories.

Built by theluckystrike — More at zovo.one