Helmsman

A friendly label shown in the app. It doesn’t rename the real server.

The accounts your server is built on — who can sign in, who is an administrator, and the background service accounts that keep it running.

The password you use to sign in to this app. To change a Linux user’s password, use People & access above.

Pick a look — saved on this device. Auto follows your system’s light/dark setting.

Bring your own keys — they never leave this server. Enter - to remove a stored key.

Run a model on this server — free, private, and offline. Helmsman uses Ollama and picks models that fit your hardware. Local models then appear in the chat model picker just like cloud ones.

Standing orders you write for the agent — how it should behave, rules to always or never follow. Added to the start of every conversation.

What the agent has learned about your server across chats. It updates this on its own — edit or clear it anytime.

What the agent is allowed to do. Turn a tool off and the agent can’t use it at all. Green tools are read-only and run without asking; the rest need your approval (except in Auto mode).

The folder new agent chats start in — usually where your server’s files and compose stacks live.

Little AI watchmen that check your server on a schedule — even when the app is closed — and report to the 🔔 bell. They only ever look (read-only): they never change anything on their own. Add an ntfy topic to also get a push to your phone.

Helmsman’s App Store is served from an online catalog so new apps appear without updating Helmsman. Point this at your own JSON to add private apps. Leave blank to use only the built-in catalog.