Headless setup

LM Studio’s llmster daemon runs models without its desktop interface. It is separate from the LLMster Apple app, which connects to the daemon’s API server.

Install the daemon

Run the appropriate official installer on the host computer:

macOS or Linux
Windows

curl -fsSL https://lmstudio.ai/install.sh | bash

irm https://lmstudio.ai/install.ps1 | iex

These commands download and run LM Studio’s installer. Follow any installer instructions for making lms available in your terminal. Source: LM Studio’s headless guide.

Start and verify the host

Start the daemon.
Terminal window
```
lms daemon up
```
Download and load a model. Use the interactive commands to choose a chat model that fits your host, then check what is loaded.
Terminal window
```
lms get
lms load
lms ps
```
Start the API server. For another device on your trusted network to connect:
Terminal window
```
lms server start --bind 0.0.0.0 --port 1234
lms server status
```
Connect the client. Save the host’s real network IP and port in LLMster’s Server settings. Do not enter 0.0.0.0 as the client address.

Connect over Tailscale

Install Tailscale on the host and your Apple device, then follow Tailscale setup. The network binding above also accepts connections through Tailscale. Save the host’s Tailscale IP and server port in LLMster, and keep authentication enabled.

Keep the service available

The daemon and the API server have separate status checks. If the client cannot connect, inspect lms server status, the host firewall, and whether the host is asleep. A running daemon alone does not prove the API server is listening.

For automatic startup, use LM Studio’s Linux service guide or its documented desktop background-service settings. See the CLI reference for lms daemon and model-management commands.

To stop serving requests:

lms server stop

The server command reference describes port and binding options.