# Claude Desktop Setup for Voice Commands
## Quick Setup Instructions
### 1. Add MCP Configuration
Edit your Claude Desktop configuration file:
```bash
# macOS
open ~/Library/Application\ Support/Claude/claude_desktop_config.json
# Windows
notepad %APPDATA%\Claude\claude_desktop_config.json
# Linux
nano ~/.config/Claude/claude_desktop_config.json
```
### 2. Add Voice Commands Server
Replace the contents with:
```json
{
"mcpServers": {
"claude-voice-commands": {
"command": "python",
"args": [
"-m",
"mcp_server.voice_server"
],
"cwd": "/Users/justinwitcoff/Test/136775783/claude-voice-commands",
"env": {
"PYTHONPATH": "/Users/justinwitcoff/Test/136775783/claude-voice-commands",
"VOICE_BACKEND": "elevenlabs",
"VOICE_FALLBACK": "google",
"LOG_LEVEL": "INFO"
}
}
}
}
```
**⚠️ Important**: Replace the `cwd` and `PYTHONPATH` with your actual project path!
### 3. Restart Claude Desktop
Completely quit and restart Claude Desktop for the changes to take effect.
### 4. Test the Integration
Try these commands with Claude Code:
#### Basic Tests:
- **"Test my microphone setup"**
- **"What voice recognition engines are available?"**
- **"List my audio devices"**
- **"Calibrate my microphone"**
#### Voice Command Tests:
- **"I want to give you a voice command"**
- **"Transcribe what I'm about to say"**
- **"Listen for my voice input for 10 seconds"**
### 5. Expected Behavior
When working correctly, you should see:
- Claude Code can access voice commands tools
- Microphone tests show your available devices
- Voice recognition uses ElevenLabs Pro for high accuracy
- Graceful fallback to Google/Whisper if needed
### 6. Troubleshooting
#### MCP Server Not Starting:
```bash
# Test server directly
cd /path/to/claude-voice-commands
source venv/bin/activate
python -m mcp_server.voice_server
```
#### Tools Not Available:
- Check Claude Desktop logs: `~/Library/Logs/Claude/mcp.log`
- Verify Python path is correct
- Ensure virtual environment is activated
#### Voice Recognition Issues:
```bash
# Test components directly
python tests/test_voice_server_direct.py
```
### 7. Advanced Configuration
You can customize the voice server by modifying environment variables in the Claude Desktop config:
```json
"env": {
"PYTHONPATH": "/path/to/claude-voice-commands",
"VOICE_BACKEND": "elevenlabs", # Primary engine
"VOICE_FALLBACK": "google", # Fallback engine
"VOICE_ENERGY_THRESHOLD": "4000", # Microphone sensitivity
"VOICE_TIMEOUT": "10", # Recognition timeout
"LOG_LEVEL": "DEBUG" # Detailed logging
}
```
## Success Indicators
✅ **MCP Server Running**: Claude Desktop starts without errors
✅ **Tools Available**: Claude Code can list voice commands tools
✅ **Microphone Working**: Test shows your devices and captures audio
✅ **Recognition Active**: ElevenLabs Pro transcription working
✅ **Voice Commands**: You can speak commands that Claude Code receives
## Next Steps
Once this is working, you can:
1. **Use Sprint 3 tools** for continuous listening mode
2. **Add wake word detection** ("Hey Claude")
3. **Create voice command macros** for common tasks
4. **Integrate with other MCP servers** for expanded functionality
---
**Need Help?** Check the test results and logs, or run the diagnostic tools in the `tests/` directory.
