Run a local model on a computer you attached with MachineSync. Token generation happens on that computer. Cloud credits are not the meter for that inference. Workspace memory still lives in The Cloud once you sign in.
Download the signed, notarized macOS app, open it, and paste the one-time pairing code. That computer becomes compute The Cloud can dispatch runs to.
The in-app steps name Ollama. Pull an open model such as gpt-oss with it.
Choose it when you start a run, or set it as the model on a Loop. Speed and context follow that computer's RAM and GPU.