Discord Voice Claude Bot

A Discord voice bot that uses OpenAI's Realtime API for natural voice conversations, with seamless integration to OpenClaw/Claude for tool-heavy requests like email, calendar, web search, and file operations.

Features

• Voice-to-Voice Conversations: Direct voice interaction using OpenAI's Realtime API (~1s latency)
• Tool Integration: Complex requests automatically routed to OpenClaw/Claude for tool access
• Natural Audio Processing: Handles Discord's Opus audio format and converts to/from OpenAI's PCM format
• Smart Function Calling: Routes requests for email, calendar, web search, commands to Claude
• Conversation Management: Reset conversation history, join/leave voice channels
• Robust Error Handling: Automatic reconnection and graceful error recovery

Architecture

Discord Voice ↔ Voice Handler ↔ OpenAI Realtime API ↔ OpenClaw Bridge ↔ Claude Tools

• OpenAI Realtime API: Handles direct voice-to-voice for casual conversation
• OpenClaw Integration: Routes complex requests requiring tools to Claude
• Audio Pipeline: Discord Opus ↔ PCM 16-bit 24kHz mono for OpenAI

Prerequisites

1. Node.js (v18 or higher)
2. Discord Bot Token with voice permissions
3. OpenAI API Key with Realtime API access
4. OpenClaw Gateway running locally at http://localhost:18789
5. FFmpeg (for audio processing) - install via:
   • Windows: winget install FFmpeg or download from https://ffmpeg.org/
   • macOS: brew install ffmpeg
   • Linux: sudo apt install ffmpeg

Setup

1. Clone and Install Dependencies

git clone <this-repo>
cd discord-voice-claude
npm install

2. Environment Configuration

Copy .env.example to .env and fill in your credentials:

DISCORD_TOKEN=your_discord_bot_token
DISCORD_CLIENT_ID=your_discord_application_id
OPENAI_API_KEY=your_openai_api_key
OPENCLAW_GATEWAY_URL=http://localhost:18789
OPENCLAW_GATEWAY_TOKEN=your_openclaw_gateway_token

3. Discord Bot Setup

1. Go to https://discord.com/developers/applications
2. Create a new application
3. Go to "Bot" section, create a bot and copy the token
4. Enable these bot permissions:
   • Scopes: bot, applications.commands
   • Bot Permissions: Connect, Speak, Use Voice Activity
5. Copy the Application ID for DISCORD_CLIENT_ID

4. OpenClaw Setup

Make sure OpenClaw gateway is running:

openclaw gateway start

Get your gateway token from OpenClaw configuration.

5. Register Discord Commands

npm run register

6. Build and Start

npm run build
npm start

Or for development:

npm run dev

Usage

Discord Commands

• /join - Bot joins your voice channel and starts listening
• /leave - Bot leaves the voice channel
• /reset - Clears conversation history

Voice Interaction

1. Use /join while in a voice channel
2. Start speaking - the bot will respond with voice for simple questions
3. For complex requests (email, calendar, web search, file operations), the bot automatically routes to Claude via OpenClaw
4. Use /reset to clear conversation history if needed

Example Interactions

Direct Voice (OpenAI Realtime API):

• "What's the weather like?"
• "Tell me a joke"
• "How are you doing?"

Routed to Claude (via OpenClaw):

• "Check my email for anything urgent"
• "What's on my calendar tomorrow?"
• "Search the web for latest news about AI"
• "Run a command to check disk space"

Technical Details

Audio Format Conversion

• Discord Input: Opus 48kHz stereo
• OpenAI Realtime: PCM 16-bit 24kHz mono (base64)
• Conversion Pipeline: Opus → PCM → Resample → Base64

OpenAI Realtime API Configuration

{
  "modalities": ["text", "audio"],
  "voice": "alloy",
  "turn_detection": { "type": "server_vad" },
  "input_audio_transcription": { "model": "whisper-1" }
}

Function Calling

The bot registers these functions with OpenAI Realtime API:

• ask_claude(message) - General Claude requests
• check_email() - Check for urgent emails
• check_calendar() - Check upcoming events
• search_web(query) - Web search
• run_command(command) - Execute commands

Troubleshooting

Common Issues

"FFmpeg not found"

• Install FFmpeg and ensure it's in your PATH

"Cannot connect to OpenClaw gateway"

• Make sure OpenClaw gateway is running: openclaw gateway status
• Check the gateway URL and token in .env

"Bot not responding to voice"

• Check Discord permissions (Connect, Speak, Use Voice Activity)
• Verify OpenAI API key has Realtime API access
• Check console logs for WebSocket errors

"Audio quality issues"

• Ensure stable internet connection
• Check Discord voice channel region/server location

Debugging

Enable detailed logging by adding to your .env:

DEBUG=discord-voice-claude:*

Check logs for:

• WebSocket connection status
• Audio processing pipeline
• Function call routing
• OpenClaw API responses

Development

Project Structure

src/
├── index.ts              # Main bot entry point
├── voice-handler.ts      # Discord voice connection management
├── realtime-client.ts    # OpenAI Realtime API WebSocket client
├── openclaw-bridge.ts    # Bridge to OpenClaw gateway
└── register-commands.ts  # Discord slash command registration

Building

npm run build    # Compile TypeScript
npm run dev      # Build and run

Adding New Functions

1. Add function definition to openClawFunctions in openclaw-bridge.ts
2. Add handler case to handleFunctionCall
3. The OpenAI Realtime API will automatically call your function when appropriate

License

MIT License - see LICENSE file for details

Support

For issues with:

• OpenAI Realtime API: Check OpenAI documentation and API status
• Discord Integration: Verify bot permissions and Discord.js version
• OpenClaw Integration: Ensure gateway is running and accessible
• Audio Processing: Check FFmpeg installation and codec support