Repolode
Submit
mbailey

mbailey/voicemode

An MCP voice layer that lets you have spoken conversations with Claude Code and other MCP-capable agents, using cloud or local speech services.

View on GitHubWebsite
Stars
1,410
Forks
198
Open issues
63
Last commit
4 days ago

What is it?

VoiceMode adds voice input and output to Claude Code and other agents that support the Model Context Protocol. You speak to the agent and hear its replies, so you can keep working when your hands or eyes are busy, such as while walking or cooking.

It can run with OpenAI's speech APIs as a cloud option, or fully locally with Whisper.cpp for speech-to-text and Kokoro for text-to-speech. The local services expose the same API as OpenAI, so VoiceMode can switch between them. Recording stops automatically when you finish speaking.

Installation is available as a Claude Code plugin or through a Python installer built on uv. It supports Linux, macOS, Windows under WSL, and NixOS, with Python 3.10 to 3.14.

Key features

  • 01Speech input and spoken replies inside Claude Code and other MCP agents
  • 02Local speech-to-text with Whisper.cpp and text-to-speech with Kokoro
  • 03Cloud fallback through the OpenAI API
  • 04Automatic silence detection to end recording
  • 05Optional audio saving for debugging

Use cases

  • Talking to a coding agent while walking or away from the keyboard
  • Keeping voice data on your machine by running Whisper and Kokoro locally
  • Reviewing agent output by listening instead of reading the screen

Quick start

  1. 1

    Add the VoiceMode marketplace to Claude Code

    claude plugin marketplace add mbailey/voicemode
  2. 2

    Install the VoiceMode plugin

    claude plugin install voicemode@voicemode
  3. 3

    Install the CLI and local voice services from inside Claude Code

    /voicemode:install
  4. 4

    Start a voice conversation

    /voicemode:converse

Things to know

  • Needs a computer with a working microphone and speakers.
  • Cloud speech services require an OpenAI API key; local services need extra setup and system dependencies such as ffmpeg and PortAudio.

Topics

#voice#mcp#claude#speech-to-text#text-to-speech#whisper#ai-agent#anthropic#asr#claudecode

FAQ

What is voicemode?
An MCP voice layer that lets you have spoken conversations with Claude Code and other MCP-capable agents, using cloud or local speech services. VoiceMode adds voice input and output to Claude Code and other agents that support the Model Context Protocol. You speak to the agent and hear its replies, so you can keep working when your hands or eyes are busy, such as while walking or cooking.
Is voicemode free to use? What license does it have?
voicemode is open source under the MIT license. Check the license terms before using it in commercial products.
Is voicemode still maintained?
Actively maintained. The last commit to voicemode was on Oct 17, 2026.
What can I use voicemode for?
Talking to a coding agent while walking or away from the keyboard Keeping voice data on your machine by running Whisper and Kokoro locally Reviewing agent output by listening instead of reading the screen
How do I install voicemode?
Add the VoiceMode marketplace to Claude Code: claude plugin marketplace add mbailey/voicemode Install the VoiceMode plugin: claude plugin install voicemode@voicemode Install the CLI and local voice services from inside Claude Code: /voicemode:install Start a voice conversation: /voicemode:converse
Where can I get voicemode?
The source code and installation instructions are on GitHub: https://github.com/mbailey/voicemode

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects