A hands-off anime AI companion that sees your screen, hears your voice, and takes the controls — a fully autonomous assistant living on your Mac.
YumiClaw isn't a chatbot in a box. It perceives your machine and acts on it directly — closing the loop between what you say and what gets done.
Captures and understands your screen with Claude vision plus local OCR — across any monitor you choose.
Wake-word activation, AI voice detection that ignores noise, Whisper transcription and lifelike spoken replies.
17 real actions — click, type, scroll, drag, shortcuts, clipboard, app launching — driving your Mac for you.
Chains multi-step tasks, verifies its own work, retries on failure, and remembers your preferences.
Say “Yumiko” to summon it — no buttons, no clicks, completely hands-free.
Silero voice detection knows real speech from background noise.
High-quality ElevenLabs TTS with playful emotes — it feels alive.
Action → verify → retry. If something fails, it tries another way.
“Open YouTube, search music, play the first video” — all in one go.
Reads and writes your clipboard for instant copy-paste workflows.
Pick exactly which display YumiClaw should watch and operate on.
Tell it once you use Chrome or Spotify — it remembers next time.
A warm worker means ~2.2s replies with no cold-start lag.
Every command flows through a tight perception-to-action pipeline — and YumiClaw closes the loop on its own.
“Yumiko”
VAD
Whisper
Claude
Vision/OCR
17 actions
TTS
Real commands, real follow-through — no copy-pasting instructions, no manual clicking.
YumiClaw is open source and runs entirely on your own machine. The full install guide is being polished — check back soon.
Want to be first to run it? Stay tuned — full setup steps land here shortly.
YumiClaw watches, listens, and works alongside you — so you can speak your intent and let it take the controls.
Get started locally