GPU-CLI - Run Code on Remote GPUs
Run cloud GPUs from your terminal
One command to spin up a GPU, run your job, and sync results back. Across RunPod, Vast and more.
Free to install. Zero markup on compute. Auto-stop when idle.
The simplicity of a CLI, the power of the cloud
Purpose-built to save you time and money
∞
Job Persistence
Close your laptop or go to sleep. Your jobs keep running. Reconnect anytime.
Auto-stop kicks in 5 minutes after your job finishes, never mid-run. Say goodbye to forgotten pods.
Skip the 30-45s wait for public IPs. Connect via relay almost instantly.
Choose your own hardware
From RTX 5090 to H100. Pay your provider directly at their rate, no markup from us.
RTX 5090
LoRA training, image generation
A40
Production inference, large models
A100 PCIe
LLM fine-tuning, transformers
H100 PCIe
Frontier research, massive models
*Approximate RunPod pricing shown for comparison. Actual rates vary by provider. View current pricing
Your code, your keys, your data
Credentials in OS keychain
Your GPU provider API key stays in your system keychain. Never stored in config files or transmitted to us.
SSH-secured transport
Code syncs directly to your GPU provider over SSH or secure tunnels. On RunPod, data at rest relies on provider encryption; supported providers can also use GPU CLI-managed LUKS.
Automatic result sync
Output files sync back to your machine as they're created. No manual downloads needed.
Config-driven setup
Reads your pyproject.toml or gpu.jsonc. Define dependencies once, run anywhere.