More time near your baby, not your keyboard.
Answer Claude Code without touching the keyboard.
Baby Coding watches for permission prompts, reads the options out loud, and lets you answer with a hand gesture or your voice. You never leave the window you are working in.
Free
How it works
Three steps, no window switching.
Baby Coding connect in your coding assistant and lives in the menu bar. It only turns the camera on when a decision is actually waiting for you.
- 01
The assistant asks
Claude Code hits a permission prompt or a multiple choice question and pauses.
- 02
Baby Coding speaks up
A floating glass panel appears above everything with the request and the numbered options, and reads them out loud so you can keep your eyes on your work.
- 03
You answer with a gesture
Thumbs up to allow, thumbs down to deny, N fingers to pick option N. Hold for about two seconds to confirm. The session resumes and the panel disappears.
Gestures
The whole vocabulary fits on one hand.
Hold a gesture inside the on-screen target for about two seconds to confirm. The HUD shows a progress bar so you always know when it registered.
- Thumbs up Yes, or option 1
- Thumbs down No, deny the request
- 1 to 5 fingers Pick option N
- Shaka Hold to dictate by voice
- Wave Cancel and close
- Pinch Open and navigate the wheel
Or just say it
Say "baby โฆ" to wake the microphone and answer out loud. Speech is transcribed on device, and the options are read back to you before anything is sent.
“baby coding run the tests”
Integrations
Built for Claude Code first.
Baby Coding is assistant agnostic by design. The others are on the way.
-
Claude Code Ready -
Cursor Soon -
Codex Soon -
opencode Soon
Safety
A gesture should never be able to wreck your repo.
Convenience only counts if a mistaken thumbs up cannot do real damage. Three guarantees make that true.
Denylist
Dangerous patterns are never auto approved by a gesture. They fall through to the assistant own dialog, where you answer with the keyboard on purpose.
Camera on demand
The camera stays off until a decision is actually pending, and shuts down right after. Gesture recognition runs entirely on your Mac.
Nothing leaves the machine
No video, no audio and no transcript is uploaded. Speech is transcribed on device, and the app never sees an API key.
rm -rf git push --force :(){ :|:& };: More time near your baby,
not your keyboard.
Install it, grant the permissions, and answer your next prompt without reaching for the mouse.
Download for macOSFree