Skip to content

Settings overview

Settings is one window with a left rail of tabs and a panel on the right. The rail is sorted into five groups by purpose, so you can find a control by asking “is this about the AI’s brains, how the app looks, the public reception desk, things I’m plugging in, or my workspace and account?”

How to get here

Open Settings from the app’s main menu or the gear icon. It opens as a modal over whatever you were doing — close it (top-right Close, or click outside) and you’re back where you were.

The five groups

🧠 Brains — the AI behind your workers

Everything that decides how your workers think.

  • Providers — connect the AI services (OpenAI, Anthropic, Gemini, and seven more) that power your workers. Also holds web-search keys, the autonomy model, and per-tool call limits.
  • Models — browse and install small AI models that run inside the app, with no external service and no per-use cost.
  • Visuals — choose which image and video generators the Designer worker reaches for, and in what order.

🎨 Appearance — how the app looks and sounds

Personalisation. None of this changes what your workers can do — only your experience of the app.

  • Theme — colour scheme (dark/light/system), text size, the Dashboard’s header animation, the window-frame animation, and an optional time-of-day auto-switch.
  • Voice — read replies aloud (text-to-speech), dictate with your microphone, and a live hands-free Talk mode.

🖥️ Reception — the public-facing desk

For running the app as a visitor-facing kiosk.

  • Kiosk — turn the app into a fullscreen touch-screen receptionist that greets walk-up visitors, with optional voice and a phone hand-off.

🧩 Integrations — things you plug in

Connections to the outside world. These tabs are documented in their own areas; this overview lists them so you know where they live.

  • MCP servers — third-party tool servers (Notion, Gmail, Slack, or your own).
  • Connectors — save an API token once so workers can reach a service without the key ever appearing in chat.
  • Channels — inbound/outbound messaging (Email and Signal today; more coming).
  • Peers — link other askTheodor machines together for sync and delegation.

🗂️ Workspace — your account, data, and housekeeping

Account-level and maintenance settings.

  • General — preferred providers, telemetry, computer-use, and resetting the workspace.
  • License — your trial status, activating a key, and the device this install is tied to.
  • Updates — your installed version and installing new releases.
  • Cost — token spend by day/week/month, plus budgets.
  • Trash — restore or permanently delete soft-deleted items.
  • Achievements — the trophy case of what you’ve unlocked.
  • Backups — export your whole workspace, or import it onto another machine.
  • Demos — preconfigured starter packs, plus factory reset.
  • Background — install an OS-level service so scheduled work keeps running while the app is closed.

In practice

The five groups answer “where would this live?” When hunting for a control, ask which question it belongs to — how do workers think (Brains), how does the app look (Appearance), the walk-up desk (Reception), what am I plugging in (Integrations), my data and account (Workspace) — and you’ll land on the right tab without scanning the whole rail.

Only three tabs matter on day one. Providers so workers can think, Theme so the app is comfortable to look at, and License when your trial runs out. Everything else waits until you have a reason.

Two tabs are worth revisiting. Cost is the only place that tells you what your workforce is actually spending, and Backups is the thing you’ll wish you’d used before a big change. Check both before letting anything run unattended.

Terms in this page

  • Worker — a configured AI character you talk to and assign work; each has its own model, tools, and memory.
  • Provider — an AI service (cloud or local) that supplies the “thinking” your workers do. You connect at least one before the app is useful.
  • Model — a specific AI brain a provider offers (e.g. a particular OpenAI or local model). One provider can offer several models.
  • Designer — the built-in worker that turns written briefs into images and videos.
  • Token — the unit AI models read and write in (roughly ¾ of a word). Providers bill per token, so several screens track them.
  • MCP server — an external program, following the Model Context Protocol standard, that hands your workers extra tools (like reading a Notion page or sending a Slack message).
  • Connector — a saved credential for an outside service (e.g. an API token) that workers use indirectly, so the secret never shows up in a conversation.
  • Channel — a way messages flow in and out of the app, like email or a messaging app.
  • Peer — another askTheodor installation, on another machine, linked to this one.
  • Kiosk — a locked-down, fullscreen mode that turns the app into a public touch-screen receptionist.
  • Telemetry — optional, local-only usage statistics (kept on your machine) that help you tune things like which model size runs well.
  • Modal — a window that opens over the current screen and must be closed to return to what’s behind it.