Call its name, say what you want — it does the whole thing.From quick chores to real coding — by voice.
or in Terminal: brew install --cask mvcompany8500-web/tap/sasarisa

Dictation apps turn speech into text. You still have to click, open, search and send — by hand.
Sasarisa finishes the job. Everyday things happen on your Mac in about a second; anything that needs thinking goes to Claude Code, which looks at the screen and does it. It gets faster the more you use it.
Green happens on your Mac in about a second. Purple goes to Claude Code.
“Play lofi on YouTube”
“Next video”
“Shop for wireless earbuds on Amazon”
“Search for MacBook reviews”
“3 minute timer”
“Remind me in 20 minutes to stretch”
“Volume 30”
“Close all tabs”
Select text, then
“Make this more polite”
“Tell Claude to fix the login bug”
ClaudeSeveral at once and your own routines work too — and it's Korean-first, with 12 languages supported.
Rename it if you like. After the first call, keep talking for 30 seconds without the name.
Talk after the chime. It acts within 0.9 seconds of you finishing.
Like Handy: listens while held, runs on release. Most accurate in noisy rooms.
Cut in any time — even over speakers, if you talk close to the mic.
Time from the end of your sentence to action (measured on a Mac mini).
What you say often moves to the instant side the more you use it.
A voice in a video must never drive your Mac. Anything irreversible waits for your yes, and your voice never leaves the Mac.
Apple silicon Mac (M1 or later) · macOS 14+ · a paid Claude plan (Pro, Max or Team)
Follow the setup guide, then run “claude” once in Terminal to log in.
One line in Terminal — or “Download for macOS” above (notarized by Apple).
Allow the microphone, speech recognition and accessibility, and download the transcription model. When ◉ shows in the menu bar, say “Sasarisa”.
Pro only covers what runs on our servers. Nothing that is free today will become paid.