Using your own local models

If you run models yourself in LM Studio, Ripul Host offers them as models for Claude Code. They appear in the model picker on your Mac and on your phone, and a chat on one works like any other: it can read the project, change files and ask for approval.

A model on this Mac

1. Start the server in LM StudioOpen LM Studio and turn on its local server. Ripul Host finds it by itself while the server is running.
2. Open the model pickerEach model LM Studio has is listed, labelled as running in LM Studio on your Mac. Loaded models come first.
3. Start a chat on itChoose a folder and a local model as you would any other. Claude Code has to be installed on the Mac, because a local model works through it.

A model on another machine

1. Open Coding Agents in Ripul HostIn Ripul Host on your Mac, open Settings, then Coding Agents, and find Local Models.
2. Add ServerGive it a name, and its address as LM Studio shows it. Add a key if that server asks for one.
3. Pick one of its modelsIts models join the list with the server’s name beside them, so the same model on two machines is easy to tell apart.

• Your Mac makes the calls, so only the Mac needs to reach the other machine. Your phone does not.
• A key stays on the Mac. It is never sent to your phone.
• A model that says it cannot use tools is marked in the list. It may not manage work on files or commands.
• A long conversation is tidied up at your model’s own limit, which is usually smaller than Claude’s.

Keeping everything on your own machines

A local model keeps the thinking on your own hardware. A Direct connection keeps the conversation off Ripul’s servers. Choose a local model in a Direct chat and nothing goes to Ripul or to an AI company.