runpod-cli

CLI for managing RunPod GPU cloud infrastructure. Optimized for AI agent workflows.

Features

• List pods with status, GPU type, region, cost, and uptime
• Start and stop pods on demand
• Create pods with network volume attachment and SSH key injection
• Delete pods with safety confirmation
• List available GPU types with full pricing (secure, community, spot)
• Filter GPUs by minimum VRAM
• List and create network volumes
• Self-describing AI manifest via --ai flag
• JSON output mode on all list commands
• Native fetch only - no external HTTP dependencies

Installation

npm install -g runpod-cli

Setup

export RUNPOD_API_KEY=your_api_key_here

Get your API key from: https://www.runpod.io/console/user/settings

Usage

Pods

# List all pods
runpod-cli pods

# List as JSON
runpod-cli pods --json

# Start a stopped pod
runpod-cli pod:start <pod-id>

# Stop a running pod
runpod-cli pod:stop <pod-id>

# Create a new pod
runpod-cli pod:create \
  --name my-comfyui \
  --gpu "NVIDIA RTX A6000" \
  --region EU-CZ-1 \
  --volume 52t492qes6 \
  --image runpod/comfyui:latest \
  --disk 50

# Create pod with SSH access
runpod-cli pod:create \
  --gpu "NVIDIA GeForce RTX 3090" \
  --ssh-key "ssh-rsa AAAA..." \
  --cloud COMMUNITY

# Delete a pod (requires --confirm)
runpod-cli pod:delete <pod-id> --confirm

GPUs

# List all GPU types with pricing
runpod-cli gpus

# Filter by minimum VRAM
runpod-cli gpus --vram 24

# Sort by VRAM descending
runpod-cli gpus --sort vram

# Find affordable high-VRAM GPUs
runpod-cli gpus --vram 48 --sort price

# Get JSON for parsing
runpod-cli gpus --json

Volumes

# List all network volumes
runpod-cli volumes

# Create a new volume
runpod-cli volume:create \
  --name comfyui-models \
  --size 150 \
  --region EU-CZ-1

Global Options

# Pass API key directly (overrides env var)
runpod-cli pods --token rp_xxxxxxxxxxxx

# Any command supports --token
runpod-cli pod:start <id> --token rp_xxxxxxxxxxxx

AI Integration

This CLI is self-describing for AI agents:

# Full JSON schema (all commands, options, types)
runpod-cli --ai

# Brief one-liner per command
runpod-cli --ai brief

# Usage examples for all commands
runpod-cli --ai examples

# Schema for a specific command
runpod-cli --ai pod:create
runpod-cli --ai gpus

Claude Code / MCP Usage

Add to your Claude configuration to give AI agents full RunPod control:

{
  "mcpServers": {
    "runpod": {
      "command": "runpod-cli",
      "args": ["--ai"]
    }
  }
}

Or in a CLAUDE.md:

Use `runpod-cli --ai` to discover available RunPod management commands.
RUNPOD_API_KEY is set in the environment.

pod:create Options Reference

| Option | Default | Description | |--------|---------|-------------| | --name | auto-generated | Pod name | | --gpu | required | GPU type ID (from runpod-cli gpus) | | --region | none | Data center ID (e.g. EU-CZ-1) | | --volume | none | Network volume ID to attach at /workspace | | --image | runpod/comfyui:latest | Docker image | | --disk | 20 | Container disk size (GB) | | --cloud | SECURE | SECURE or COMMUNITY | | --ports | 8188/http,8080/http,22/tcp | Exposed ports | | --ssh-key | none | Public SSH key (injected as PUBLIC_KEY env) |

License

MIT