rp — RunPod CLI

A CLI for managing RunPod GPU pods. Two tiers: low-level (rp pod create) for bare pods, and opinionated (rp up) for fully configured pods with tools, secrets, auto-shutdown, and remote Claude support.

Install

uv tool install https://github.com/agencyenterprise/rp.git
rp --install-completion  # optional tab completion

Quick Start

# Store your API key in macOS Keychain
rp secrets set RUNPOD_API_KEY

# Create a fully configured pod (tools, secrets, auto-shutdown)
rp up h100

# Run commands on it
rp run h100-1 -- nvidia-smi

# Launch remote Claude
rp claude h100-1 -p "Run the training script in /workspace/project"

# Check progress
rp status h100-1

# Sync logs and destroy when done
rp down h100-1

Commands

Managed Pods (opinionated)

Command	Description
rp up [template] [--gpu SPEC --storage SIZE] [--alias NAME] [--network-volume ID]	Create pod + install tools + inject secrets + deploy auto-shutdown
rp claude <alias> [-p PROMPT] [-d DIR]	Launch Claude in tmux on pod (autonomous or interactive)
rp status <alias>	Check remote Claude progress and read report
rp logs <alias>	Sync remote Claude logs locally
rp down <alias> [--skip-logs]	Sync logs and destroy a pod (counterpart to rp up)
rp setup <alias>	Re-run pod setup (recovery from partial failures)
rp secrets list|set|remove|inject	Manage secrets in macOS Keychain; inject pushes to a running pod

Connection & Execution

Command	Description
rp run <alias> -- <cmd>	Execute command on pod via SSH
rp shell <alias>	Interactive SSH shell
rp code <alias> [path]	Open VS Code remote SSH
rp scp <src> <dest>	Copy files to/from pod via SCP

Low-Level (rp pod)

Command	Description
rp pod create [template] [--gpu --storage ...] [--alias]	Create bare pod, run setup.sh
rp pod start <alias>	Resume stopped pod (re-injects secrets for managed pods)
rp pod stop <alias>	Stop pod
rp pod destroy <alias> [-f]	Terminate pod permanently
rp pod track <pod_id_or_name> [alias]	Track existing pod by ID or name
rp pod untrack <alias>	Stop tracking (doesn't terminate)
rp pod list	List all pods with status
rp pod show <alias>	Detailed pod info
rp pod clean	Remove invalid aliases and orphaned SSH config
rp pod gpus [-f FILTER]	List available GPU types (e.g. -f 'vram>=80')

Templates

Built-in: h100, 2h100, 4h100, 5090, a40 (all 500GB container disk, no volume, using {project}_{person}_{i} naming).

# Set naming variables in .rp_settings.json (or ~/.rp_settings.json for global)
echo '{"project": "ast", "person": "alex"}' > .rp_settings.json

rp pod create h100      # creates ast_alex_1, ast_alex_2, etc.

# Override per-command
RP_PROJECT=other rp pod create h100   # creates other_alex_1

# Custom templates
rp template create ml --alias-pattern "{project}_{person}_{i}" --gpu 2xA100 --storage 1TB
rp template list
rp template delete ml

Configuration

Settings are defined in .rp_settings.json files at any directory level (walks cwd → root, closest wins). See docs.md for details.

{"person": "alex", "project": "ast", "secrets": ["HF_TOKEN", "WANDB_API_KEY"]}

Additional config in ~/.config/rp/:

• pods.json — aliases, metadata, templates
• setup.sh — script for bare pods (customizable)