Getting Started with the TensorPool CLI

A comprehensive guide to managing your GPU infrastructure from the command line with our powerful CLI tool.

Tycho Svoboda Cofounder & CEO

September 27, 2025

5 min read

The TensorPool CLI is your command-line interface to powerful GPU infrastructure. Whether you're spinning up a single H100 for model training or orchestrating a multi-node cluster for distributed workloads, the CLI makes it fast and simple.

Video Tutorial

Watch our comprehensive video guide to get started with the TensorPool CLI:

TensorPool Tutorial 1: Spin Up Your First GPU TensorPool

Prerequisites

Before getting started, make sure you have the following:

  • A TensorPool account (sign up at tensorpool.dev)
  • Python 3.8 or later installed
  • SSH keys generated (we'll show you how)

Step-by-Step Guide

Step 1: Install the TensorPool CLI

Install the CLI using pip:

pip install tensorpool

Verify the installation:

tp --version

Step 2: Set Up Your API Key

Get your API key from the TensorPool Dashboard and set it as an environment variable:

your_api_key_here

TIP: Add this to your ~/.bashrc or ~/.zshrc to make it persistent.

Step 3: Generate SSH Keys

If you don't have SSH keys yet, generate them:

ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519

Press Enter to accept the defaults when prompted.

Step 4: Create Your First Cluster

Let's create a single-node cluster with one H100 GPU:

tp cluster create \
  -i ~/.ssh/id_rsa.pub \
  -t 1xH100 \
  --name my-first-cluster

For a multi-node cluster with 8 B200s per node:

tp cluster create \
  -i ~/.ssh/id_rsa.pub \
  -t 8xB200 \
  -n 4 \
  --name distributed-training

Available instance types: 1xH100, 2xH100, 4xH100, 8xH100, 1xH200, 8xB200, and more.

Step 5: List and Connect to Your Cluster

View all your clusters:

tp cluster list

This shows cluster ID, instance type, SSH username, IP addresses, ports, and hourly pricing. SSH into your cluster using the provided details:

ssh tensorpool@192.168.1.42

Step 6: Create and Attach NFS Storage

Create a 500GB NFS volume for persistent storage:

tp nfs create \
  --s 500 \
  --name nfs-test

List all NFS volumes:

tp nfs list

Attach the NFS volume to your cluster:

tp nfs attach <storage_id> <cluster_ids>

Step 7: Tear Down Resources

When you're done, clean up resources to stop billing:

# Detach NFS from cluster
tp nfs detach <storage_id> <cluster_ids>

# Destroy NFS volume
tp nfs destroy <storage_id>

# Destroy cluster
tp cluster destroy <cluster_id>

WARNING: Destroying a cluster or NFS volume is permanent. Make sure to back up any data before running destroy commands.

Common Use Cases

💻 Single-GPU Development

Quick prototyping and model experimentation:

tp cluster create -i ~/.ssh/id_rsa.pub -t 1xH100 --name dev

🚀 Multinode Training

Large-scale model training across multiple nodes:

tp cluster create \
  -i ~/.ssh/id_rsa.pub \
  -t 8xB200 \
  -n 2 \
  --name llm-training

📊 Data Processing Pipeline

Attach shared NFS storage for data pipelines:

# Create cluster
tp cluster create -i ~/.ssh/id_rsa.pub -t 4xH100 --name pipeline

# Create and attach 1TB NFS
tp nfs create --size 1000 --name data-store
tp nfs attach --cluster-id <id> --nfs-id <nfs_id>

Next Steps

  • Check out the full CLI documentation on GitHub
  • Join the TensorPool Slack to ask questions and share your projects
  • Read our blog post on optimizing multi-node GPU clusters

Ready to get started? Sign up for TensorPool and get your first cluster running in minutes.