For Mac mini and Mac Studio
Plan across the Macs you use.
A screenless Mac can run your models while you check its status from the Mac you work on. LMM host mode, paired views, and planning across Macs are planned capabilities.
- Planned — host mode and paired views have not shipped.
- Automatic login is required until no-login host support is resolved.
01 · Host
Let the mini or Studio do the serving.
The planned host is a background service with no window or menu bar. It starts through launchd when the host account logs in. Until support for running without a logged-in user is settled, turn on automatic login on the host.
The planned setup routes are the LMM app on the host, an installer over SSH, or the bundled command-line tool. None is available yet; there is no command to run today.
02 · Pairing
Pair once, then connect over TLS.
The planned flow starts on the host. It shows a short pairing code; you enter that code once on the Mac you use. After the devices exchange keys, their host connection uses TLS.
- On the same network, Bonjour is intended to find the host.
- For another network, you can enter a hostname or IP address.
- A Tailscale hostname is one example you could enter. Tailscale is not an LMM integration, relay, or guarantee that the connection will work.
03 · Access
Reading status and changing things are separate.
Pairing is intended to allow the paired Mac to read host status. Actions such as loading or unloading a model need a separate per-device control permission that you grant on the host.
04 · Planner
Choose where each task should run.
The planned multi-Mac Planner compares the Macs you pair and can place work on the machine suited to that task. It is intended to show the endpoint settings for each agent. Cross-Mac planning is a roadmap capability and is not available yet.
05 · Privacy
Prompts go to the model server you choose.
That server receives a prompt to generate its response. It could run on this Mac, a paired Mac, or a remote endpoint you configured. TLS protects the planned connection to a paired host; it does not change where the model runs.
LMM does not upload prompt content for telemetry. The planned observer records conversation sizes and timings locally by default. Saving message content is a separate option and is off by default.
LMM also makes network calls for the knowledge feed, licence activation and renewal, update checks, and paired-host connections.
Availability
Host mode is still planned.
Trial downloads are not available yet, and purchasing opens at launch.
Trial access and purchasing open at launch.