What to set up after your first chat, in the order most people need it, from cheaper models to backups.

The Quickstart gets you chatting with one model. These are the next things worth setting up, roughly in the order most people want them. Everything here is in Settings, at the bottom of the left sidebar; Settings pages lists them all.

Spend less on routine work

Suzent runs small jobs in the background, such as titling chats and saving memories. By default they use your main model. Set a cheaper, faster model as the Cheap role in Settings → Model Roles, and those jobs use it instead. See Model roles.

To see what you're spending, open Settings → Usage, or type /status in a chat.

Keep your work safe

Choose a default permission mode. The mode you pick on the new-chat screen, before sending anything, becomes the default for new chats. You can still change it in any chat. See Permissions & approvals.
Run code in a sandbox. With Docker installed, turn the sandbox on in Settings → Security so commands run in an isolated container instead of on your computer. See Using the sandbox.

Teach it about you

Tell the agent about yourself and your work in a chat; it remembers across conversations. Check what it has kept in the Memory panel and correct anything wrong. See Memory.

For things you do often, add a skill so it follows the same steps every time.

Reach it from anywhere

From your phone or another computer: see Desktop, phone, and browser.
From a chat app such as Telegram or Slack: see Chat apps.
Keep it running after you close the window, so chat apps and scheduled tasks keep working: turn on Settings → Background Service. See Automation.

Back it up

Turn on GitHub Sync in Settings → Data & Sync to keep your settings, skills, and memory in a private repository you own. It also moves them to a new computer. API keys are never synced.

Look and language

Settings → Appearance switches between English and 简体中文, light or dark theme, and the accent colors.

Making it yours