Skip to main content

Getting Started

Table of Contents​

Instance lifecycle

Launch an instance​

Choose the framework of your choice and configure the instance with

  • GPU type (A30, L4, A100, A100-80GB, H100, H200, RTX-PRO6000)
  • Storage - Scale up to 2TB of storage. Need more? Contact us.
  • Startup script - Run your own startup script to install any library or package you want.

and hit launch. We do all the required work for you, in less than few minutes your instance is up and running.

info

Instances are charged per minute. You can pause and resume the instance anytime you want. Once you pause the instance, the compute is released and can be used by other users. You will be charged only for the reserved storage.

Accessing an instance​

Once you launch an instance, you can access it in multiple ways.

  • JupyterLab
  • VS Code
  • API - You can run services like Gradio, Streamlit, FastAPI, Flask, etc on port 6006, and access it using the API endpoint. On templates that come with their own web UI (ComfyUI, Automatic1111, Fooocus), it opens that UI.
  • SSH - You can SSH to the instance using the provided DNS. Ensure you have added your public key here.

JupyterLab, VS Code and a template's web UI ask for your instance password. See Instance password below.

All the GPU powered instances comes with required Cuda libraries and required libraries pre-installed. You can also install any library or package you want using the terminal.

info

Always store the data under /home, any data stored outside /home will be lost when you pause and resume the instance.

System-level libraries installed outside /home (for example, via apt-get or global pip install) are lost when you pause and resume the instance. To persist your dependencies, install them into a conda environment or a Python venv under /home — both survive pause/resume. See Environments for details, or let the jl run CLI manage a venv for you automatically.

Instance password​

Every instance has one password, and it works the same way on every template. It opens:

  • JupyterLab and VS Code, on all templates.
  • The template's web UI, opened from the API button, on ComfyUI, Automatic1111 and Fooocus.

To find it, select Password on your instance in the dashboard, then the eye icon to show it or the copy icon to copy it. It is also on the instance's Access tab.

  • Refreshing a page keeps you signed in. A new tab, or reopening the browser, asks again.
  • Every resume makes a new password. Copy it again from the dashboard after a resume. A restart keeps the same password.
  • SSH uses your SSH keys, not this password.
  • Services you run yourself on port 6006 (Gradio, FastAPI, vLLM), and the Ollama template's API, are not covered by the password and stay public.
  • Keep it private: anyone with the password can use your instance.
  • Instances started before this update have no Password button: JupyterLab opens directly from the dashboard, and VS Code and the web UI work as before. If they ask for a password, run cat /home/.jarvis_password in a JupyterLab terminal. Restart, or pause and resume, once to switch to the dashboard password.

To call a template's web UI from code, send the password as a bearer token:

curl -H "Authorization: Bearer <instance password>" "<API endpoint>/..."

Pause an Instance​

You can pause your instance by clicking the pause button. Pausing an instance frees up compute resources like GPU, CPU, and RAM. All the data will be retained and will be available when you resume.

Any libraries installed outside a custom conda environment will be lost.

info

Data only under /home will persist when you pause an instance, always use /home for storage.

You will still be charged for the storage, so if you do not plan to use the instance anytime soon, consider deleting it.

Resume an Instance​

Click resume on a paused instance to start with your work. A new instance will be created with all your previous data. You can also modify the instance parameters like GPU type, GPU quantity and storage while resuming the instance.

info

You can resume an instance only if it is in the same region.

Delete an Instance​

Make sure you have backed up all the required data before deleting the instance. Once you delete the instance, all the data will be lost and cannot be recovered.

CLI & Python SDK​

All of the above — launch, pause, resume, delete — can also be done from the terminal using the jl CLI. It works with AI coding agents like Claude Code, Codex, Cursor, and OpenCode.

jl create --gpu A100 --storage 50
jl pause <machine_id>
jl resume <machine_id>
jl destroy <machine_id>

See the full CLI docs and Python SDK docs for complete reference and examples.