Tailscale setup

LLMster works with Tailscale, so you can reach your models from another Wi-Fi network or a cellular connection. Tailscale connects your devices through a private network called a tailnet. Your computer still runs the model and needs to stay awake and online.

Connect your devices

1. Install Tailscale on the host and your Apple device. Follow the Tailscale quickstart and sign in to the same account on both devices. They should appear in the same tailnet. On iPhone or iPad, allow Tailscale to add its VPN configuration and turn the connection on.

2. Start LM Studio’s API server. Follow LM Studio setup, including Serve on Local Network and Require Authentication. Keep the token for LLMster’s Auth Key field. For a headless host, follow Headless setup and start the server with:

   lms server start --bind 0.0.0.0 --port 1234

   If it’s already running with a localhost-only binding, stop it with lms server stop before restarting with these options. See the LM Studio server command reference.

3. Copy the host’s Tailscale IPv4 address. Find the computer running LM Studio in Tailscale’s device list or Machines page. Use that computer’s Tailscale address, not the phone’s address or the host’s home-network IP. Tailscale’s connection guide explains where to find it.

4. Save it in LLMster. Open Settings → Server. Enter a Server URL such as http://100.101.102.103:1234, replacing the example IP and port with your host’s values. Use only the base URL, without /v1 or another API path. Paste your LM Studio token into Auth Key, then tap Test Connection. A Tailscale sign-in or auth key doesn’t belong in this field.

5. Try it away from your home network. Keep Tailscale connected on both devices. On an iPhone with cellular data, turn Wi-Fi off, test the connection again, and send a short chat message. On a Wi-Fi-only device, try another Wi-Fi network.

HTTP over Tailscale

The URL above uses HTTP, but traffic between your Tailscale devices travels through an encrypted tunnel. Tailscale keeps it encrypted even if a relay is needed. This protection applies to the Tailscale connection; an ordinary HTTP connection to your home-network IP doesn’t gain encryption from this setup. See Tailscale’s encryption overview.

Keep access private

Keep LM Studio authentication enabled. Serve on Local Network and --bind 0.0.0.0 can also make the API reachable on your LAN. Allow the server’s TCP port through the host firewall for your intended clients, and check that your tailnet’s access rules permit your Apple device to reach it.

You don’t need router port forwarding, an exit node, or a subnet router when Tailscale is installed on both devices. This setup uses the host’s private Tailscale address and doesn’t require publishing the API with Tailscale Funnel.

If it won’t connect

• Confirm both devices are connected to the same tailnet and signed in. Reauthenticate if Tailscale asks.
• Check that the host is awake and the LM Studio API server is running with network access enabled.
• Check the host’s Allow incoming connections setting, firewall, and tailnet access rules for the server port.
• Recheck the host’s Tailscale IP, port, and LM Studio token. A home-network address such as 192.168.1.20 won’t use this Tailscale connection.
• If another VPN is active on your Apple device, check for a conflict and confirm Tailscale stays connected during the test.

Connection troubleshooting More checks for server addresses, authentication, and network permissions.