Hey Product Hunt, I am the maker of Sarah. I built it because desktop voice assistants still feel disconnected from actual work. I wanted something that stays out of the way: press a key, speak, keep working. The three workflows are dictation into any Mac app, screen-aware AI commands with current app/window/URL/screenshot/OCR context, and quick voice Q&A in a floating overlay. The hard product problem is trust. A useful desktop voice assistant needs sensitive permissions: Microphone, Accessibility, Input Monitoring, and sometimes Screen Recording. Sarah handles that by being open source, showing permission state clearly, and making agent/runtime setup explicit instead of hidden. I would love feedback on whether the hotkey model is understandable, whether the permission flow feels clear enough, and what Mac workflows this should support next.

Sarah

A macOS voice assistant for dictation and AI commands

A macOS voice assistant for dictation and AI commands