ToolPiper sets most of itself up. On its first launch it downloads what voice and tool calling need with no click, and the Get started card on the Home pane walks you through the rest one step at a time. The one choice only you make is a language model. Once Choose a language model shows a checkmark, you can ask ToolPiper questions, and the steps left are ones you can finish whenever you like.

The steps

  • Bring your models comes first when ToolPiper finds models from another app on your Mac, such as Ollama, and offers to use them where they are, with no copy and no download. It does not appear otherwise.
  • Set up voice and Set up tool calling install themselves on first launch and show their progress while they do. Voice is speech recognition and speech output; tool calling is the model that picks the right tools for your requests.
  • Choose a language model is yours: use a model already on your Mac, add a cloud provider, or download the one ToolPiper recommends for your Mac. See Choosing a language model.
  • Grant permissions lists the macOS permissions ToolPiper's features use and walks you through allowing them. It is open from the start, and choosing a language model adds the permissions an agent's tools need.
  • Try dictation opens once voice is ready, and Ask ToolPiper something opens once a language model is ready.
  • Sign in to ModelPiper comes last, after the other steps, and only if you are signed out.

What done looks like

Each finished step shows a checkmark, and the card counts them, such as "3 of 7". A step that is waiting says what it waits for, and a step that is installing shows its progress. Once the steps are done the card offers Dismiss, and the line under the Home title says ToolPiper is running on this Mac. Signing in is never required to dismiss the card.

What you can do once it is set up