Claude Listen Plugin

Voice input for Claude Code using listen CLI tool with local Whisper transcription.

Features

• Voice input instead of typing
• Multi-language support (es, en, fr, de, it, pt, zh, ja, ko)
• Multiple Whisper models (tiny to large)
• MCP server integration with slash commands

Requirements

• Claude Code
• listen CLI
• Node.js v18+
• Working microphone

Installation

1. Install listen CLI

git clone https://github.com/gmoqa/listen.git
cd listen
pip install -r requirements.txt
# Ensure 'listen' is available in PATH

2. Install plugin in Claude Code

Open Claude Code and run:

/plugin marketplace add gmoqa/listen-claude-code
/plugin install claude-listen@gmoqa/listen-claude-code

That's it! The plugin will automatically:

• Install MCP server dependencies
• Configure the listen voice tool
• Add the /listen command

Usage

/listen command

/listen

Starts voice recording. Press Ctrl+C when done speaking. Claude processes the transcription automatically.

Direct MCP tool

Claude can also use the listen_voice tool directly when needed.

How it works

1. User runs /listen
2. MCP tool calls listen CLI
3. User speaks, then presses Ctrl+C
4. Whisper transcribes audio to text
5. Claude processes the text as a normal request

Testing

cd test
./quick-test.sh    # Fast validation
./auto-test.sh     # Full test suite

Troubleshooting

"listen command not found" Add listen to PATH or set LISTEN_PATH in .mcp.json

"No module named 'whisper'" pip install -r requirements.txt in listen directory

Microphone not working Check permissions in System Preferences → Privacy → Microphone

License

MIT - See LICENSE file

Credits

• listen by @gmoqa
• Whisper by OpenAI
• Claude Code MCP integration