AI导哈

语音输入MCP服务器

一款用于语音输入的MCP服务器，可录制语音笔记或将语音指令交给Claude处理，适合在Claude Code中以语音方式完成输入和指令操作。

  • by rabb1tpt
  • Python
  • 本地部署
  • 更新于
在 GitHub 查看 rabb1tpt/MCP-voice

使用说明

MCP-voice

An MCP server for voice input - record voice notes or speak commands for Claude to process.

Setup

1. Create virtual environment

cd ~/Bruno/code/MCP-voice
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Configure in your project

Add a .mcp.json file in your project root:

{
  "mcpServers": {
    "voice": {
      "command": "/path/to/MCP-voice/.venv/bin/python",
      "args": ["/path/to/MCP-voice/server.py"],
      "env": {
        "VAULT_DIR": "/path/to/your/vault"
      }
    }
  }
}

3. Restart Claude Code

The voice tools will be available.

Environment Variables

Variable Description Default
VAULT_DIR Path to the vault (must have an inbox/ directory) . (current directory)
WHISPER_VENV Path to Whisper virtual environment ~/code/openai-whisper/.venv
WHISPER_MODEL Whisper model to use base

Tools

Tool Description
record_voice_note Record, transcribe, and save to vault inbox
listen Record, transcribe, and return as input for Claude to process

Usage

In Claude Code:

  • "record Bitcoin thoughts" - saves transcription to inbox
  • "listen" - speak your request, Claude processes it

Dependencies

  • arecord (ALSA) for audio recording
  • Whisper for transcription (configured via WHISPER_VENV)

相关 MCP