Run VLAs, VLMs, world models, MuJoCo, and Isaac Lab

The easiest way to build with robot foundation models

Start in a browser tab with free MuJoCo demos. Scale to cloud GPUs when the model, simulation, or training job needs more than your machine.

Launch robot models like local tools

Run a local daemon and use the same API shape when you later switch to hosted VLAs, VLMs, world models, or GPU-backed simulation.

$ runrobot pull php-parkour-g1
$ runrobot run php-parkour-g1 --browser
Loading MuJoCo WASM scene...
Loading custom JS policy runner...
✓ G1 parkour is running locally

Start local. Scale with cloud.

Browser and local execution are free. Cloud usage applies when RunRobot hosts GPU inference, GPU simulation, or training for you.

Free locally. Paid when GPUs work for you.

We keep the public plans simple, with Ollama-style cloud usage tiers. The metering engine tracks normalized GPU time behind the scenes.

Free

Browser and local usage

$0
Download
  • Unlimited local and browser MuJoCo runs
  • Public model manifests
  • Free browser demos for G1 parkour and IIWA + Sharpa
  • Upgrade prompts for GPU cloud models

Pro

50x more cloud usage than Free

$20 / mo

or $200/year

Get Pro
  • Everything in Free
  • 3 cloud models, sims, or training jobs at a time
  • Private model uploads
  • Day-to-day VLA/VLM/world-model work

Max

5x more usage than Pro

$100 / mo
Get Max
  • Everything in Pro
  • 10 cloud jobs at a time
  • Priority capacity
  • Heavy sustained robot learning

Your robot data stays yours

Local runs stay on your hardware. Cloud prompts, observations, actions, checkpoints, and simulation artifacts are not used for training unless you explicitly opt in.