Skip to content
Simulo

Simulo

Cloud-first, simulation-first robot learning. Write a task in Python, run it on a GPU in the cloud, watch the policy train — all from the simulo SDK and CLI.
Get StartedCLI ReferenceAPI ReferenceOpen Console

Simulo treats learning as the primary abstraction: environments, robots, sensors, policies, training, evaluation, and visualization compose into one reproducible, cloud-executed workflow. You author a task in Python; Simulo runs it on GPU-backed infrastructure and streams back logs, checkpoints, trained models, and recordings.

Install

Section titled “Install”

The simulo package is a torch-free SDK and CLI with no GPU or platform requirement: it authors, submits, and observes jobs from any machine running Python 3.11+. The jobs themselves run on Simulo’s cloud GPUs.

Terminal window
pip install simulo
simulo login

The lifecycle

Section titled “The lifecycle”

  1. Install — pip install simulo gets you import simulo and the simulo CLI from one package.

  2. Create — simulo create my_task scaffolds a runnable app that already pins the validated global cartpole.

  3. Login — simulo login opens a browser, signs you into the Simulo cloud, and stores your credentials locally.

  4. Run — simulo run my_task/app.py submits the job to the Simulo cloud: it packages your app, uploads it, and runs it on a GPU worker, streaming logs back to your terminal.

  5. Observe — simulo logs, simulo view (live 3D viewport), simulo result, and simulo outputs take you from a running job to its saved output.

  6. Make it yours — browse another ready-to-use catalog asset, or validate and publish your own USD/URDF package.

  7. Continue — simulo run my_task/app.py --from <job-id> picks up training from an earlier job’s checkpoint instead of starting over.

See Getting Started for the full walkthrough, and Core Concepts for how the pieces fit together.

Simulo global catalog

Section titled “Simulo global catalog”

The Simulo global catalog is the fastest way to put a robot, world, or prop into a job. It is a read-only collection of validated, versioned assets that you reference directly from the CLI or Python—no USD download or copy into your app:

Terminal window
simulo asset list --global
simulo asset inspect simulo/robot/cartpole:v1
cartpole = simulo.Asset.from_registry("simulo/robot/cartpole:v1")

The catalog currently includes:

CartpoleFranka PandaHumanoidUnitree H1JetBotPick-and-placeANYmal CWarehouse worldLab tableMounting stand

Start with Choose & validate assets to browse global assets, read validation results, and learn when to publish into your organization’s catalog instead.

Explore the docs

Section titled “Explore the docs”
Getting StartedInstall Simulo, choose a validated global-catalog asset, and run your first cloud job.
Core ConceptsApps, jobs, tasks, runtimes, versioned asset catalogs, storage, callbacks, and observability.
GuidesTask-oriented walkthroughs: your own robot, evaluation, recordings, live viewing, and more.
ExamplesComplete apps you can copy, save, and run — from a CPU hello job to GPU training, evaluation, manipulation, and live-viewed scenarios.
Simulo ConsoleBrowse assets and jobs, inspect logs and output, watch a live run, or cancel work from your browser.
CLI ReferenceEvery simulo command and flag, including the complete nested asset lifecycle.
API ReferenceThe public Python API — App, Runtime, Asset, Volume, callbacks, and the Task and Scenario contracts.