| name | setup-handy |
| description | Install and configure Handy, the free, local, open-source push-to-talk voice dictation app, so the user can voice-code in any terminal. Use when someone wants to set up voice input / dictation / "talk to Claude", asks to "set up Handy" or "set up voice coding", or is starting the Lifestyles of the AI-Native workshop. Detects macOS vs. Windows and walks through install, model download, permissions, and a hotkey test. |
Set up Handy (voice coding)
Handy is a free, open-source, fully-local push-to-talk dictation app. Transcription runs on-device (Whisper / Parakeet) — nothing leaves the machine, and there's no subscription. That makes it the right pick for a room of developers: no card, no cloud, works offline on conference Wi-Fi. This skill gets the user dictating in under five minutes, then proves it works.
Voice coding is the on-ramp for the whole workshop: hands off the keyboard, talk at 184+ wpm, and drive several agents at once. Handy is just the dictation layer — Claude Code is what listens.
Procedure
-
Detect the platform. Run uname -s (Darwin = macOS, otherwise check for Windows). Pick the matching install path below.
-
Install Handy.
-
macOS (Homebrew preferred):
brew install --cask handy
If Homebrew isn't present, point them to the signed .dmg at https://handy.computer and ask them to drag Handy to Applications.
-
Windows: download the installer from https://handy.computer and run it. (Confirm with the user once it's installed.)
-
First launch + model download. Open Handy. On first run it downloads a local speech model — recommend Whisper Small (fast, accurate enough for code) or Parakeet if offered. Wait for the download to finish before testing.
-
Grant permissions (macOS). Handy needs Microphone and Accessibility permission to type into other apps:
- System Settings → Privacy & Security → Microphone → enable Handy.
- System Settings → Privacy & Security → Accessibility → enable Handy.
Tell the user exactly where to click; this is the most common place setup stalls.
-
Set the push-to-talk hotkey. In Handy's settings, confirm or set a comfortable push-to-talk key (hold to talk, release to transcribe). A modifier like right-Option or F5 works well. Tell them the chosen key.
-
Prove it works. Ask the user to focus this terminal, hold the hotkey, and say a test sentence such as "Claude, list the files in this directory." Confirm the dictated text appears in the prompt. If nothing types: re-check Accessibility permission (step 4) — that's the culprit ~90% of the time.
-
Report. Tell the user Handy is live, name their hotkey, and hand off: "You're voice-coding now — hold {hotkey}, talk, release. Try driving me hands-free."
Notes
- Don't recommend a paid tool here. Handy is deliberately the choice because it's free and local — no hard sell mid-workshop.
- If
brew install --cask handy fails (cask not found in their tap), fall back to the .dmg from https://handy.computer rather than fighting Homebrew.
- Everything is local: reassure privacy-conscious users that audio never leaves their machine.