A comprehensive OpenClaw skill for managing GPU cloud instances on CFGPU platform. Includes GPU Sniper for auto-detecting and grabbing available GPU resources.
- GPU Sniper: Auto-poll all regions and GPU types, grab resources when available
- Resource Probe: One-shot scan showing availability table across all regions/GPUs
- Resource Management: List available regions, GPU types, and system images
- Instance Lifecycle: Create, start, stop, release GPU instances
- Image Management: List and manage system/user images
- Interactive Wizard: User-friendly command-line interface
- API Integration: Full CFGPU API coverage
clawhub install cfgpu-apigit clone https://github.com/r600a-code/GPUCLI.git
cd GPUCLI
chmod +x scripts/*.shNo manual configuration needed. Just run and the script will prompt for your API token on first use.
# Quick probe — see what's available right now
./scripts/probe.sh
# Auto sniper — keep polling and grab when available
./scripts/gpu-sniper.sh./scripts/probe.shScans all region × GPU combinations and outputs a table:
RTX 4090 H800 L40S A100-PCIe
杭州 无货 无货 无货 无货
重庆 无货 无货 无货 无货
绍兴 ✅ 有货 无货 无货 无货
./scripts/gpu-sniper.sh # 默认30秒轮询
./scripts/gpu-sniper.sh --interval 10 # 10秒轮询
./scripts/gpu-sniper.sh --gpu-num 2 # 抢2卡When a GPU becomes available:
═══════════════════════════════════════
🎉 抢到了!
GPU: RTX 4090 x1
区域: 绍兴
实例ID: instance-xxxxx
价格模式: Day | 时长: 1
═══════════════════════════════════════
| Script | Description |
|---|---|
gpu-sniper.sh |
GPU 抢卡 — 持续轮询,有卡自动创建实例 |
probe.sh |
资源探测 — 一次性扫描,输出可用性表格 |
cfgpu-helper.sh |
Main CLI tool for all operations |
setup-env.sh |
Interactive environment setup |
check-config.sh |
Configuration validation |
example-usage.sh |
Usage examples |
This skill supports the full CFGPU API:
- ✅ Region management
- ✅ GPU type listing
- ✅ Image management (system/user)
- ✅ Instance lifecycle (create/start/stop/release)
- ✅ Instance status monitoring
- ✅ Paginated instance queries
- ✅ GPU resource detection and auto-grab
- Never commit API tokens to version control
- Use environment variables or secure token files
- Set appropriate permissions on token files (
chmod 600) - Regularly rotate API tokens for security
- Fork the repository
- Create a feature branch
- Make your changes
- Submit a pull request
- WeChat:
wwwr600a - Twitter/X: @ADfunAI
MIT License - see LICENSE file for details.