Table of contents

Launch a Pod

This is the full click-by-click flow for launching a QuickPod from the customer portal. The launch wizard has three steps — Instance Type, Template, and Configure — shown as a progress indicator at the top of the page.

Before you start

You need:

  • A verified account — launching a pod requires KYC verification. If you aren’t verified, a verification prompt appears instead of the wizard. See KYC.
  • A project to launch into. See Creating Projects.
  • Wallet balance — pods bill per second while running. See Wallet & Billing.
  • (Optional) An SSH key if you want shell access. Keys can only be added during pod creation, so add one first at SSH Keys if you don’t have one. See SSH Keys.

Step 1 — Select Instance Type

  1. Go to Pods and click Launch Pod.
  2. On the Select Instance Type step, browse the available GPUs. Use the search box (Search GPUs (e.g., A100, RTX, H100)…) to filter.
  3. By default only GPUs with live availability are shown. Tick Show All GPUs to include everything, including sold-out types.
  4. Each GPU card shows its VRAM, CUDA/Tensor cores, the included CPU and RAM (e.g. Includes 4 vCPU + 16GB RAM), an availability badge (e.g. 3/8 Available), and the price per minute and per hour.
  5. Click Select on the GPU you want.
    • If the GPU is sold out, the button reads Join Waitlist instead. Click it to queue for capacity; you’ll be notified when it frees up. See Troubleshooting.

Selecting a GPU auto-fills sensible defaults: 1 GPU, 100% of the card, and the CPU/RAM included with that GPU.

Fractional GPUs. You don’t have to take a whole card. You’ll set the exact GPU count and the percentage of each GPU in Step 3 — see Fractional GPUs.

Step 2 — Choose a Template

  1. On the Choose a Template step, either:
    • Click a 1-click template to start from a pre-configured app (image, ports, and command already set), or
    • Click Custom Configuration to configure everything manually.
  2. Filter templates with the search box (Search templates by name, image, GPU…) or the All Categories dropdown. Templates you’ve saved yourself appear under the User Templates category.
  3. Each template card shows its name, description, container image, and resource chips (vCPU, RAM, GPU type). Click one to pre-fill the configuration and jump to Step 3.

For the full catalog and how to save your own, see Templates.

Step 3 — Configure Your Pod

The Configure Your Pod step is a set of cards on the left with a live Summary panel on the right.

Basic Info

FieldNotes
Pod NameRequired. A descriptive name, e.g. my-training-pod.
ProjectRequired. The project this pod is billed to.
Container ImageThe Docker image. Pre-filled from your template, or defaults to a PyTorch image for custom configs.
Private Registry (requires authentication)Tick to reveal Registry URL, Registry Username, and Registry Password/Token for private images.

Resources

  • vCPU and Memory (GB) — set each with the slider or the number box. When a GPU is selected, the amount included with it is free; anything above that adds cost. A live caption shows how much capacity is free right now.
  • GPU Type — read-only; shows the GPU you picked in Step 1.
  • Number of GPUs — how many GPUs to attach (1 up to what’s available).
  • GPU Memory/Cores per GPU — the percentage of each GPU to use, from 5% to 100% in 5% steps. This is the fractional-GPU control. A donut chart shows what’s already booked, your request, and what’s free, and a summary line shows the effective VRAM.

NFS Volume (optional)

Attach persistent storage so your work survives restarts:

  1. In the NFS Volume card, pick a volume from the list (or No Volume for ephemeral storage only).
  2. Set the Mount Path Inside Pod — defaults to /data, the path every Podstack image expects.
  3. No volume yet? Click Create Volume (a project must be selected first).

See Storage & Data for details.

Configuration

  • SSH Key Access — pick an SSH key to enable shell access. Port 22 is exposed automatically. SSH keys can only be added during pod creation, so add one first if the list is empty. Choose No SSH Access to skip.
  • Environment Variables — click + Add Variable to add KEY / value pairs.
  • Entry Point Command (optional) — override the container’s default command, e.g. python app.py.
  • Exposed Ports — enter a port and click Add Port to make it reachable over HTTPS. Each container port is mapped to a public URL after launch. Jupyter/notebook images default to port 8888.

Billing Period

Choose how you’re billed: Per Minute, Per Hour, Per Day, Per Month, Per Quarter, or Per Year. Longer periods can carry a discount — click Compare Savings to see the breakdown. The default is Per Minute.

Review and launch

  1. Check the Summary panel — it shows your GPU allocation, CPU, memory, billing period, and the total cost with GST.
  2. Click Launch Pod. (The button stays disabled until Pod Name and Project are set.)
  3. You’re taken to the Pods list, where the new pod appears and moves through creating → pending → running.

Tip: Want to reuse this exact setup? Click Save as Template in the Summary panel before launching. See Templates.

If you see an insufficient balance prompt, top up your wallet and try again. If the GPU became unavailable between Step 1 and launch, you’ll be offered the waitlist.

Connect to your pod

Open the pod from the Pods list to reach its detail page. Depending on how you configured it, you’ll see:

Web Terminal

On a running pod, click Open Terminal in the Web Terminal card for an interactive browser shell — no SSH client or keys required.

SSH

If you attached an SSH key, the SSH Access card shows a ready-to-copy command:

ssh -i ~/.ssh/<key_name> podstack@ssh-<subdomain>.cloud.podstack.ai

Replace <key_name> with your private key file. Click Configure SSH for step-by-step setup on macOS, Linux, Windows PowerShell, or Windows Git Bash — first-time connections need a one-time SSH config entry. See SSH & PowerShell.

Jupyter / Notebook

On notebook images, the Notebook Access card shows the Notebook URL, Username (default podstack), and an auto-generated Password. Copy the password now — it isn’t shown again after you leave the page.

Exposed ports

The Exposed Ports & Endpoints card lists every port and its public URL:

  • HTTP services: https://<subdomain>-<port>.cloud.podstack.ai
  • SSH (port 22): ssh-<subdomain>.cloud.podstack.ai

Click Edit to add or remove ports on a running pod. See Manage & Scale.

Next steps