Skip to main content
Bed of Roses (BOR) is an AI companion for macOS — not a chat window. A small, draggable presence floats above your real desktop, always one keystroke away. You tell it what you want, and it does the work in front of you. Bed of Roses working live on the desktop: streaming its reasoning, writing files into editor cards, and running commands while the result takes shape

​
What a conversation can do

Every request can end in something real, not just text:
  • An app — “make me a calorie tracker” opens a native Mac window, built for you, saved forever. Building apps
  • Action in your services — with Gmail, Calendar, Slack, Notion, GitHub and ~100 more connected, “reply to Sarah” uses the real Gmail, not a browser. Connectors
  • Web work — BOR drives a real Chrome with its own profile: reads pages, fills forms, stays signed in, verifies results. Ambient browser
  • File work — it reads PDF, Word, Excel, HEIC, and almost anything else; writes and edits files; every path it mentions clicks through to Finder. Files
  • Mac control — screen capture, vision-guided clicks and typing, opening apps. Computer use
  • Media — YouTube, TikTok, Instagram, tweets, and video play as cards inside the presence. Media
  • Scheduled work — “every morning at 8…” runs real agent turns while you’re away and notifies you. Background tasks
  • A new look — eight themes, generated wallpapers applied to your actual desktop. Themes
And you watch all of it happen: the thought bubble streams BOR’s reasoning, the code it writes, the commands it runs, and the pages it visits as live cards — with a stop button that pauses any task and a play button that resumes it. The presence

​
Yours, locally

  • Your model. Anthropic, OpenAI, Google Gemini, or any OpenAI-compatible endpoint (DeepSeek, Kimi, Z.ai, Ollama presets) — or Bed of Roses credits with no key at all. Switch models from a chip in the input bar. Providers
  • Your machine. Everything runs locally. Keys stay in local config, generated apps are sandboxed, and every action lands in an audit log.
  • Your memory. It remembers you across sessions — and you can run several presences, each a fully separate companion.

​
Get started

  1. Installation — npm install -g bed-of-roses, then bor onboard.
  2. Onboarding — name it, style it, connect a model, link your services.
  3. Quickstart — your first useful requests.