JarvisLabs

JarvisLabs.ai is a GPU cloud for AI development that lets you launch on-demand GPU and CPU instances (H100, H200, A100, RTX Pro 6000, A6000, A5000, L4, A30) from the terminal. Its Python SDK (jarvislabs / legacy jlclient) and jl CLI wrap an API for the full instance lifecycle - create, pause, resume, and destroy - plus GPU type discovery, framework templates, persistent filesystems, and managed runs for training and inference workloads, billed per minute of compute.

5 APIs 0 Features
AIGPUCloudInfrastructureCompute

APIs

JarvisLabs Instances API

On-demand GPU and CPU instance lifecycle - create, list, get, pause, resume, rename, and destroy instances - exposed through the jarvislabs Python SDK and jl CLI, with per-minut...

JarvisLabs GPU Types API

Lists available GPU types with real-time availability, region, VRAM, RAM, vCPU, and per-GPU-hour pricing (surfaced via the jl gpus command), used to pick hardware before launchi...

JarvisLabs Templates API

Lists available framework templates (PyTorch, TensorFlow, JAX, VM) and persistent filesystems used to provision instances with a ready environment.

JarvisLabs Inference & Deploy API

Managed runs that upload code, build a virtual environment, install dependencies, run a training or inference script on a fresh or existing instance, and auto-pause on completio...

JarvisLabs SDK & CLI

The jarvislabs Python SDK and jl CLI (and the legacy jlclient library) wrap the JarvisLabs API for instance management, GPU discovery, templates, filesystems, SSH keys, startup ...

Collections

Pricing Plans

Jarvislabs Plans Pricing

3 plans

PLANS

Rate Limits

Jarvislabs Rate Limits

4 limits

RATE LIMITS

FinOps

Resources

👥
GitHubOrganization
GitHubOrganization
🔗
LinkedIn
LinkedIn
🔗
Website
Website
🔗
Documentation
Documentation
🔗
Plans
Plans
🔗
RateLimits
RateLimits
🔗
FinOps
FinOps

Sources

Raw ↑
opencollection: 1.0.0
info:
  name: JarvisLabs API
  version: '1.0'
request:
  auth:
    type: apikey
    key: Authorization
    value: '{{apiToken}}'
    in: header
items:
- info:
    name: Instances
    type: folder
  items:
  - info:
      name: List instances
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/instances
    docs: Returns all instances for the authenticated account. Equivalent to User.get_instances() in the SDK and `jl list`
      in the CLI.
  - info:
      name: Create an instance
      type: http
    http:
      method: POST
      url: https://api.jarvislabs.ai/v1/instances
      body:
        type: json
        data: '{}'
    docs: Launches a new GPU or CPU instance. Equivalent to Instance.create() in the SDK and `jl create` in the CLI. Persistent
      storage survives pause/resume; billing is per minute of compute.
  - info:
      name: Get an instance
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/instances/{machine_id}
    docs: Returns details for a single instance. Equivalent to User.get_instance(machine_id) and `jl get <machine_id>`.
  - info:
      name: Rename an instance
      type: http
    http:
      method: PATCH
      url: https://api.jarvislabs.ai/v1/instances/{machine_id}
      body:
        type: json
        data: '{}'
    docs: Renames an instance. Equivalent to `jl rename <machine_id> --name`.
  - info:
      name: Destroy an instance
      type: http
    http:
      method: DELETE
      url: https://api.jarvislabs.ai/v1/instances/{machine_id}
    docs: Permanently deletes an instance and all of its data, stopping all billing. Equivalent to instance.destroy() and
      `jl destroy <machine_id>`.
  - info:
      name: Pause an instance
      type: http
    http:
      method: POST
      url: https://api.jarvislabs.ai/v1/instances/{machine_id}/pause
      body:
        type: json
        data: '{}'
    docs: Pauses a running instance, stopping compute billing while persistent storage is retained. Equivalent to instance.pause()
      and `jl pause <machine_id>`.
  - info:
      name: Resume an instance
      type: http
    http:
      method: POST
      url: https://api.jarvislabs.ai/v1/instances/{machine_id}/resume
      body:
        type: json
        data: '{}'
    docs: Resumes a paused instance, optionally switching GPU type, GPU count, or storage and attaching a filesystem. Equivalent
      to instance.resume() and `jl resume <machine_id>`.
- info:
    name: GPU Types
    type: folder
  items:
  - info:
      name: List GPU types
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/gpus
    docs: Lists available GPU types with real-time availability, region, VRAM, RAM, vCPU count, and per-GPU-hour pricing.
      Equivalent to `jl gpus`.
- info:
    name: Templates
    type: folder
  items:
  - info:
      name: List templates
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/templates
    docs: Lists available framework templates used to provision instances. Equivalent to User.get_templates() and `jl templates`.
- info:
    name: Filesystems
    type: folder
  items:
  - info:
      name: List filesystems
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/filesystems
    docs: Lists persistent storage volumes. Equivalent to FileSystem.list() and `jl filesystem list`.
  - info:
      name: Create a filesystem
      type: http
    http:
      method: POST
      url: https://api.jarvislabs.ai/v1/filesystems
      body:
        type: json
        data: '{}'
    docs: Creates a new persistent storage volume. Equivalent to FileSystem.create() and `jl filesystem create --storage`.
  - info:
      name: Delete a filesystem
      type: http
    http:
      method: DELETE
      url: https://api.jarvislabs.ai/v1/filesystems/{fs_id}
    docs: Deletes a persistent storage volume. Equivalent to FileSystem.delete() and `jl filesystem remove <fs_id>`.
- info:
    name: Account
    type: folder
  items:
  - info:
      name: Get account balance
      type: http
    http:
      method: GET
      url: https://api.jarvislabs.ai/v1/account/balance
    docs: Returns the authenticated account's wallet balance. Equivalent to User.get_balance() and `jl status`.