What is VoiceLayer?¶
VoiceLayer lets Claude talk to you and listen.
When you use Claude Code, everything happens through text — you type, Claude types back. VoiceLayer adds voice to that conversation.
What Can It Do?¶
- Claude speaks to you through your speakers — status updates, explanations, questions
- You speak back through your microphone — Claude transcribes what you said and continues working
- Everything stays local — your voice is transcribed on your machine, nothing leaves your computer
Real Examples¶
QA Testing a Website¶
You're reviewing a client's website with Claude. Instead of typing descriptions of what you see:
Claude (voice): "How does the checkout page look on mobile?"
You (voice): "The payment form is cut off on the right. And the submit button is hidden behind the keyboard."
Claude records this, moves to the next page, and keeps testing.
Hands-Free Code Review¶
You're walking through code changes while Claude takes notes:
Claude (voice): "I found three changes in the auth module. Want me to walk through them?"
You (voice): "Yes, start with the middleware changes."
Background Notifications¶
Claude finishes a long task while you're reading docs in another window:
Claude (voice): "Build complete. 47 tests passing, 2 skipped."
No need to switch back to the terminal to check.
How It Works (Simply)¶
- You add VoiceLayer to Claude Code (one line in a config file)
- Claude gains 5 new voice tools — announce, brief, consult, converse, and think
- When Claude wants to speak, it calls the voice tool
- When Claude needs your input, it speaks a question, records your answer, and reads the transcription
The entire flow happens in your terminal. No browser, no app, no account needed.
What You Need¶
- A Mac or Linux computer with speakers and a microphone
- Claude Code installed
- Bun (a JavaScript runtime — one command to install)
- sox (for microphone recording — one command to install)
- edge-tts (for text-to-speech — one command to install)
- whisper.cpp (for speech-to-text — optional but recommended for fully local operation)
Total setup time: about 5 minutes. See the Quick Start guide.
Is It Free?¶
Yes. VoiceLayer is open source (Apache-2.0 license). All voice processing runs locally on your machine. The only optional cloud component is Wispr Flow for speech-to-text, which requires an API key — but the default whisper.cpp backend is fully local and free.
VoiceLayer is part of the Golems ecosystem — three open-source MCP servers for memory (BrainLayer), voice (VoiceLayer), and orchestration (cmuxLayer).