Voice access to Claude — just say "Hey Claude" to activate, like Siri or Alexa. Dictate tasks, ask questions, and send prompts to Claude Code hands-free.

How it works

Hey Claude runs entirely on your machine. It uses OpenAI Whisper locally via faster-whisper — your voice never leaves your computer.

When you press the shortcut (or say the wake word), Hey Claude starts listening. It detects when you stop speaking, transcribes what you said, and pastes the result directly into the Claude Code input — then sends it automatically. No clicks, no typing.

Features

Wake word activation — say "Hey Claude" to activate, just like Siri or Alexa. No keyboard needed.
Long dictation mode — say "start dictation" to enter a buffered session where everything you say is accumulated, then sent all at once when you say "end dictation"
Automatic send — in wake word mode, your prompt is sent automatically after 3 seconds of silence
Stop on command — say "stop" at any time to exit voice input
Fully configurable — wake word, stop phrase, dictation phrases, and language are all customizable from VS Code settings
Works with accents — handles pronunciation variants and imperfect speech naturally, so you don't need to speak perfectly
Hands-free — dictate tasks and questions directly into Claude Code without touching the keyboard
Status bar feedback — always-visible color-coded indicator of the current voice state

Requirements

Python 3.10+
The following Python packages (installed automatically on first use):
```
faster-whisper sounddevice pynput pyperclip numpy webrtcvad-wheels
```
Microphone access

Keyboard Shortcuts

Platform	Shortcut	Action
macOS	`Cmd+Option+V`	Toggle voice input
Windows	`Ctrl+Alt+V`	Toggle voice input
Linux	`Ctrl+Alt+V`	Toggle voice input

Pressing the shortcut a second time (or saying "stop") stops listening.

Usage

Quick command

Press Cmd+Option+V (macOS) or Ctrl+Alt+V (Windows/Linux) to start listening
Speak your prompt
Hey Claude sends it to Claude Code automatically

Wake word

Say "Hey Claude" — the status bar turns yellow and shows Waiting for Task
Speak your prompt — it is sent automatically after 3 seconds of silence
Say "stop" to exit at any time

Long dictation

Say "start dictation" — the status bar turns red and shows Dictating...
Speak freely — everything is buffered, nothing is sent yet
Say "end dictation" — the full transcription is sent to Claude Code at once

Extension Settings

Setting	Default	Description
`hey-claude.wakeWord`	`Hey Claude`	Phrase to activate wake word mode
`hey-claude.stopPhrase`	`stop`	Phrase to stop listening
`hey-claude.dictationStartPhrase`	`start dictation`	Phrase to enter long dictation mode
`hey-claude.dictationEndPhrase`	`end dictation`	Phrase to send the dictation buffer
`hey-claude.language`	`en`	Spoken language for transcription
`hey-claude.focusTarget`	`claude`	Where to focus before pasting (`claude` or `none`)

Status Bar

State	Background	Meaning
Hey Claude	—	Ready
Listening...	🟡 Yellow	Waiting for speech
Waiting for Task	🟡 Yellow	Wake word detected — speak your task
Dictating...	🔴 Red	Long dictation mode active
Thinking...	—	Processing audio

Changelog

0.1.2

Reduced silence detection window from 0.75s to 0.5s — wake word activation is now noticeably faster and more responsive

0.1.1

Added repository field to package manifest
Improved marketplace page: added How it works section, fixed status bar table, improved feature descriptions
Cleaned up extension package (excluded dev files via .vscodeignore)

0.1.0

Initial release
Wake word activation ("Hey Claude")
Long dictation mode ("start dictation" / "end dictation")
Automatic send after silence
Fully configurable wake word, stop phrase, dictation phrases, and language
Color-coded status bar feedback

Hey Claude