Jarvislabs Docs

Jarvislabs is a GPU cloud for building, training and serving AI. Launch a GPU instance with JupyterLab or VS Code, run a full VM, deploy an autoscaling LLM endpoint, and connect it all with private networking and storage. You pay by the minute, and there's nothing to set up.

Create an accountLaunch a GPUInstall the CLI

What do you want to do?

InstancesDevelop on a GPU

Pick a template with JupyterLab, VS Code or SSH. Pause it when you're done and you only pay for storage.

VMRun a full virtual machine

Root access, Docker, and your own stack, with VPC, reserved IP and security group support.

ServerlessServe an LLM endpoint

OpenAI-compatible API on vLLM, SGLang or Ollama, with autoscaling and optional scale-to-zero.

CLI & SDKAutomate from code or an agent

Create GPUs, run scripts and fetch results with jl. Works with Claude Code, Codex, Cursor and more.

AppsDeploy an app or UI

Ship FastAPI, Flask, Gradio or Streamlit apps, or run ComfyUI and Automatic1111.

Quickstart

Every option needs an account with wallet credit. Sign up, then recharge your wallet.

Dashboard

1. Open Templates and pick one, such as PyTorch.
2. Choose a GPU, storage size and region, then click Launch.
3. When the instance is running, open JupyterLab or VS Code from the dashboard, or connect over SSH.
4. When you're done, Pause the instance to stop compute billing. Your data under /home is kept. Delete it to stop all billing.

See Managing instances for the full lifecycle, or follow the hands-on Getting Started tutorial.

CLI

# Install and authenticate (also installs skills for AI coding agents)
uv tool install jarvislabs
jl setup

# Run a script on a fresh GPU — the instance is paused automatically when it finishes
jl run train.py --gpu A100 --requirements requirements.txt

# Or manage an instance yourself
jl create --gpu A100 --storage 100 --name my-experiment
jl ssh <machine_id>
jl pause <machine_id>

See the CLI reference for every command.

Python SDK

pip install jarvislabs
jl setup   # saves your API token for the SDK and CLI

from jarvislabs import Client

with Client() as client:
    # Create a GPU instance (blocks until running)
    inst = client.instances.create(gpu_type="A100", name="my-instance")
    print(f"SSH: {inst.ssh_command}")
    print(f"Notebook: {inst.url}")

    # Pause to stop compute billing (data persists)
    client.instances.pause(inst.machine_id)

See the Python SDK reference for more examples.

Choosing compute

	Instance	VM	Serverless
Best for	Interactive development, notebooks, training runs	Custom stacks, Docker, long-running services	Production LLM inference
You get	A container from a template, with JupyterLab, VS Code and SSH	A full virtual machine with root and SSH	An OpenAI-compatible HTTPS endpoint
Docker	Not supported	Supported	Managed for you
Networking	API endpoint and SSH	Public IP, plus VPC, Reserved IP, Security Groups	Managed endpoint
Scaling	Change GPU type, count or storage on resume	Pick GPU and storage at launch	Autoscales between min and max workers
Billing	Per minute while running, storage while paused	Per minute while running, storage while paused	Per running worker, plus model storage

All products

Compute

TemplatesReady-made environments: PyTorch, TensorFlow, Axolotl, Ollama, ComfyUI and more.VMGPU virtual machines with full root access.ServerlessAutoscaling, OpenAI-compatible LLM endpoints.EnvironmentsPython environments that survive pause and resume.

Networking

VPCConnect your VMs over a private network.Reserved IPKeep a VM's public IP the same across pause and resume.Security GroupsAllow-only firewall rules for VM traffic.

Storage & data

Shared FilesystemsPersistent storage you can attach to multiple instances.Object StorageS3-compatible buckets for datasets, checkpoints and weights.Download DataPull datasets and models onto your instance.

Developer tools

CLIManage GPUs and run scripts from your terminal with jl.Python SDKCreate, pause and resume instances from Python.dstackOrchestrate dev environments, training and services.

Account

Settings & billingRecharge your wallet, create API tokens and add startup scripts.TeamsShare credits and track your team's instances.

Need help?

• Troubleshooting covers common issues and fixes.
• FAQs answers billing, storage and template questions.
• Tutorials has step-by-step guides for real workloads.
• Contact us at support@jarvislabs.ai.