Hands-free voice conversations with Claude Code
Launch App opens the web UI — requires the desktop app running in the background.
Speak naturally -- local Whisper STT transcribes on-device with zero cloud latency. Auto-listen detects speech and starts recording hands-free.
Responses read aloud via local Kokoro TTS. 60+ voices, adjustable speed (0.5x--3x), paragraph-level highlighting as Claude speaks.
Watch Claude Code working in real-time. Full ANSI color, scrollback, arrow key navigation, and interactive prompt detection.
Talk (full voice), Type (type in, hear back), Read (speak in, read response), and Text (keyboard only). Switch with one tap.
macOS and Windows via Electron. Floats above all windows, draggable header, visible on all workspaces. Auto-updates from GitHub.
Content updates (UI, server logic) pulled from GitHub on every launch. Shell updates via GitHub Releases. Always current, zero effort.
Download, open, talk. That's it.
Download the app for macOS or Windows. Open it -- Murmur checks for prerequisites, starts a Claude Code session, and connects automatically. A guided tour walks you through all controls on first launch.
Tap the talk button or press Right Cmd to start recording. Speak your request, release to send. Claude's response appears as conversation bubbles and is read aloud. Or just type in the text input.
Tap the mode button to cycle between Talk, Type, Read, and Text. Each mode adjusts mic and speaker behavior. Expand the terminal panel to see Claude's full output and navigate with arrow keys.
Switch modes with a single tap on the mode button.
Grab the latest release for your platform.
git clone https://github.com/hdt325/murmur.git
cd murmur && npm install
cd electron && npm install && cd ..
npm run electronnpm i -g @anthropic-ai/claude-code
uv tool install voicemode