System Requirements
Windows
CUDA Toolkit 11.0-12.5 required
Linux
CUDA Toolkit 11.0-12.5 required
Key Features
Hardware-Aware AI Coding
Smart CUDA Intelligence
- Context-aware autocomplete for GPU kernels
- Ctrl+K editing with architecture-specific suggestions
- Chat assistant trained on CUDA optimization patterns
- Hardware-aware code generation (Ampere, Ada Lovelace, Hopper)
Real-Time GPU Profiling
Production NCU Integration
- Live performance metrics as you code
- 13+ hardware metrics (SM efficiency, memory bandwidth, occupancy)
- Color-coded performance indicators (🟢 🟡 🔴)
- One-click kernel profiling with ▶️ gutter buttons
Personalized AI Guidelines
.rightnowrules Automation
- Auto-generated guidelines based on your GPU architecture
- Personalized optimization recommendations
- Project-specific CUDA development rules
- Seamless integration with all AI features
Privacy-First AI Access
Flexible Provider Options
- OpenRouter BYOK with 200+ models
- Local models (Ollama, vLLM, LM Studio)
- Complete offline capability
- Your code never leaves your machine with local providers
Profiling Data Persistence
Optimization History Tracking
- Persistent profiling data across sessions
- Performance timeline and trend analysis
- Content-based kernel identification
- Team-shareable optimization insights
Zero-Friction CUDA Development
Get Productive Instantly
- Auto-detects your GPU and CUDA setup
- No configuration needed - just install and code
- Works with existing CUDA projects immediately
- Intelligent error messages with actionable fixes
Get Started
Quick Start Guide
Get productive with CUDA development in 3 simple steps.