runpod
par andrewharp
DevOps & CloudRésumé
Manage RunPod GPU cloud instances - create, start, stop, connect to pods via SSH and API.
Description
Ce que fait ce skill
Manages RunPod GPU cloud instances from the command line. Handles pod creation, lifecycle control, SSH access, and remote filesystem mounting via SSHFS. Works through the runpodctl CLI and RunPod's proxy infrastructure.
Handles SSH key management and SSHFS mounting that RunPod's web console doesn't support, letting agents manage the full pod lifecycle without leaving the terminal.
Quand l'utiliser
- Spinning up an RTX 4090 pod for a model training run
- Mounting a remote pod's filesystem locally to edit scripts
- Checking which GPU pods are active and their current status
- Accessing a Jupyter or ComfyUI web service running on a pod
- Sending a dataset file to a pod before starting a job
Exemple de workflow
User asks: Create an RTX 4090 pod with a 100GB workspace and connect to it via SSH
- 1
Run runpodctl create pod with --gpuType 'NVIDIA GeForce RTX 4090', --volumeSize 100, and --volumePath '/workspace'
- 2
Run runpodctl get pod to retrieve the new pod's ID
- 3
Run runpodctl start pod <id> to start the pod
- 4
Run runpodctl ssh connect <id> to get the SSH connection command
SSH command printed: ssh -p <port> root@<ip> -i ~/.ssh/runpod_key
Prérequis
- RunPod account (console.runpod.io)
- RunPod API key
- runpodctl CLI (brew install runpod/runpodctl/runpodctl)
Installation
clawhub install andrewharp/runpodExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.