Bring your own agent.
Use the AI you already pay for, or any API key. Recall is the calm workspace; the brains are your choice.
Switch per chat.
The model picker sits in the chat box. Each engine shows whether it is signed in or still needs setting up. Pick one and carry on typing.
One engine, the whole app.
The engine you make default runs chat, mail triage, your morning brief, meeting answers and voice. A meter shows how much of the chat's context is used and how much of your plan is left, with the reset time. API keys show an estimate; local models are unlimited.
The usage panel: context in this chat, plan use and the other engines.
Guided setup.
Pick an engine Recall doesn't have yet and it offers to set it up: it installs it from the official source, then opens the provider's own sign-in.
Engines that need setup, like Mistral and LM Studio, are listed in the picker.