Get started

Install with Homebrew

The tap ships the CLI/MCP package plus the notarized VoiceBar notch app as a cask. VoiceBar owns the microphone-permissioned voice daemon, so install both.

brew tap etanhey/layers
brew install etanhey/layers/voicelayer
brew install --cask etanhey/layers/voicebar
voicelayer setup

*Prefer npm? bun add -g voicelayer-mcp (or npm i -g voicelayer-mcp) installs just the CLI/MCP package.

Configure MCP

VoiceLayer runs as a singleton daemon on a Unix socket — each Claude session connects through a lightweight socat shim. Add this to any repo's .mcp.json.

// .mcp.json
{
  "mcpServers": {
    "voicelayer": {
      "command": "socat",
      "args": ["STDIO", "UNIX-CONNECT:/tmp/voicelayer-mcp.sock"]
    }
  }
}

Install prerequisites

For voice recording: sox (brew install sox). For local transcription: whisper.cpp (brew install whisper-cpp) plus a GGML model file. Cloud transcription via Wispr Flow is available as a fallback.

brew install sox whisper-cpp
# Download a model:
curl -L -o ~/.cache/whisper/ggml-large-v3-turbo.bin \
  https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v3-turbo.bin

*macOS: grant microphone access to your terminal app in System Settings > Privacy > Microphone.

Try your first voice command

Use the announce mode for a quick fire-and-forget test. The agent will speak the message aloud.

// In a Claude Code session:
voice_speak("Hello from VoiceLayer!")