Documentation

Settings & personalization

Everything about how the assistant looks, behaves and works for you lives under Settings (in the account menu). Tabs are grouped into Personal, Capabilities and System, and there’s a search box across all of them so you can jump straight to an option.

Personal

Profile

Who you are, so answers fit your time, place and priorities:

  • Your name, Timezone and Home address.
  • Priorities - what matters to you, used to judge what’s important in your Inbox.
  • Inbox rules and exact drop patterns - what should never reach your inbox.
  • Your default security mode and default memory access for new chats.

Agent

Make the assistant yours:

  • Name and Soul (a Markdown description of its personality - “friendly, concise, dry humor”). This replaces the default persona everywhere.
  • Your agents - your specialist sub-agents, each with a persona, a “when to use” hint, required capabilities (Web / Devices / Documents / Memory / Coding / Workspace), an optional surface binding and an optional default model. You can publish one to the marketplace or adopt published ones.

Appearance

Applies instantly on this device and is also saved to your account:

  • Theme (System / Dark / Dim / Light), accent color (presets or a custom picker), reading font (serif or sans), font scale (S / M / L) and density (comfortable / compact).
  • Language (System / Deutsch / English).
  • Message details - show or hide the timestamp, model name and cost under each answer.
  • Input - Enter sends, or Enter inserts a newline and Cmd/Ctrl+Enter sends.
  • Accessibility - reduce motion, high contrast.
  • Navigation - which entries appear in the sidebar, and your start page after sign-in (the main chat or any visible entry).

Behavior

How the assistant works for you:

  • New-chat defaults - the model, reasoning effort, mode (standard / coding) and data classification (standard / internal) that new chats start with.
  • Response style - answer length (concise / balanced / detailed) and answer language (match the message, or a fixed language).
  • Proactive assistant - how often it reaches out on its own (Off / Once a day / Every 6 hours / Hourly) plus its own quiet hours.
  • Notifications - the push master switch, which events notify you (proactive nudges, tool approval requests, questions from the assistant), and quiet hours. In-app cards always appear; push goes to your phone.

Capabilities

System

  • App - download the native apps (Linux desktop, Android, browser extension, terminal UI) and set up MCP access for other tools: the MCP server URL (OAuth) plus personal access tokens for headless clients.
  • Companion - the native app’s own settings, shown only inside the app: connection (background WebSocket) mode, location sharing, Health Connect metrics, location zones, and per-device sensors - see Devices.

Usage & budget

The Usage tab shows your token use and cost over a chosen range (7 / 30 / 90 / 365 days): totals for tokens, cost, runs, cache hits and tool calls, charts over time, and a per-day breakdown by model.

If a monthly budget applies to you, a bar shows how much of it you’ve spent. Budgets cascade user > org > global (the most specific limit wins); when you hit the cap, new runs are declined until next month, and the cap also stops sub-agents from spawning mid-task. Admins set budgets under Admin > Budgets (see Admin console).