Skip to main content
BOR is provider-agnostic. You bring your own model — or use Bed of Roses credits — and BOR works the same either way: the tools, the apps it builds, the browser, memory, everything behaves identically regardless of which model is behind it.

​
The five providers

You can connect several at once and switch between them from the model chip in the input bar.

​
Reasoning effort

Every provider has a reasoning effort setting — Off, Low, Medium, High (recommended), Max — controlling how much the model thinks before acting. The thinking streams into the bubble as reasoning cards. Set it per provider in onboarding or provider Settings; see Configuration.

​
Vision and images

  • Screenshot-driven features (computer use, browser screenshots, image attachments) need a vision-capable model — all current Anthropic/OpenAI/Gemini models qualify; local models vary.
  • Image generation is available with OpenAI (gpt-image-1) and Gemini (Imagen) keys. Anthropic doesn’t offer one — pair keys if you want images alongside Claude.

​
Switching

  • Click the model chip and pick any connected provider/model — instant.
  • Ask BOR: “switch to Gemini.”
  • With multi-presence, different presences can run different providers side by side.
Model ids rotate every few months. The in-app picker always reflects the live list; the ids named in these docs are a snapshot.