Voicemode

Voicemode is an MCP server that enables natural, bidirectional voice conversations with Claude Code and other Model Context Protocol clients. Built for developers and power users who want hands-free interaction, the server connects your machine's microphone and audio output to conversational speech pipelines. It supports local voice synthesis and speech-to-text processing using Whisper.cpp and Kokoro, eliminating external network dependencies for offline environments, while also allowing cloud fallback via OpenAI audio APIs. Users can initiate speech sessions directly from their terminal or coding environment, speak naturally, and receive spoken audio responses with low latency and automated silence detection. By bringing speech recognition and voice synthesis directly into agentic development workflows, Voicemode allows developers to debug, brainstorm, or conduct code reviews while away from the keyboard, reducing continuous screen time and facilitating multitasking during compute-heavy tasks or physical movement.

Category: Design, Media & Creative

Tags: audio, sound, voice, voice-changer

Visit Voicemode

How to install and configure Voicemode

1. Install the required system audio packages (for example, on macOS run brew install ffmpeg node portaudio, or use sudo apt install -y ffmpeg gcc libasound2-dev libasound2-plugins libportaudio2 portaudio19-dev pulseaudio pulseaudio-utils python3-dev on Ubuntu/Debian). 2. Install uv if needed, then run the installer package: bash uvx voice-mode-install 3. Register the MCP server in Claude Code by executing: bash claude mcp add --scope user voicemode -- uvx --refresh --from voice-mode voicemode-mcp-launcher 4. Optionally, configure your OpenAI key as a fallback for local speech models: bash export OPENAI_API_KEY="your-openai-key" 5. Start interacting using voice by running claude converse or consult the repository for Claude Code plugin instructions.

What you can do with Voicemode

• Conversing with Claude Code hands-free while away from the desk, cooking, or pacing during meetings. - Debugging and discussing software architecture orally to rest your eyes after long screen sessions. - Running local speech-to-text with Whisper.cpp and text-to-speech with Kokoro without sending audio data to cloud providers. - Dictating instructions and listening to synthesized responses during hands-busy tasks or remote terminal workflows. - Testing voice interaction loops in conversational agents using automatic silence detection and optional audio recording debugging.

Key facts

• https://github.com/mbailey/voicemode
• Design, Media & Creative
• audio, sound, voice, voice-changer

Part of MCP Servers

What is Voicemode?

Voicemode is a voice interface server built on the Model Context Protocol. It allows AI assistants like Claude Code to listen to human speech and respond aloud, providing hands-free interaction through speech-to-text and text-to-speech engines.

How do I install Voicemode?

First install system audio dependencies like portaudio and ffmpeg. Then install the Python package via `uvx voice-mode-install` and add it to Claude Code with `claude mcp add --scope user voicemode -- uvx --refresh --from voice-mode voicemode-mcp-launcher`.

Can Voicemode run completely offline?

Yes. Voicemode supports local speech-to-text using Whisper.cpp and local text-to-speech using Kokoro. These local services maintain privacy and avoid third-party cloud API costs, though OpenAI cloud APIs can also be used as fallbacks.

Is Voicemode open source?

Yes. Voicemode is distributed as an open source project under the MIT license, hosted on GitHub by user mbailey.

Which operating systems are compatible with Voicemode?

Voicemode is compatible with Linux, macOS, NixOS, and Windows native or WSL2, running on Python versions 3.10 through 3.14. WSL2 requires specific PulseAudio packages for microphone access.