mrrustybutter/discord-mcp

Discord Bot MCP

A Model Context Protocol (MCP) server that provides Discord bot functionality with voice transcription (using Google Gemini) and text-to-speech (using ElevenLabs).

Features

• 🤖 Full Discord bot functionality through MCP
• 🎤 Voice channel transcription using Google Gemini
• 🔊 Text-to-speech in voice channels using ElevenLabs
• 💬 Text channel messaging
• 📝 Transcript logging
• 🔄 Real-time voice activity detection

Prerequisites

1. Discord Bot

   • Create a bot on Discord Developer Portal
   • Enable the following intents:
     • GUILD_MESSAGES
     • MESSAGE_CONTENT
     • GUILD_VOICE_STATES
   • Get your bot token and client ID

2. Google Gemini API

   • Get an API key from Google AI Studio

3. ElevenLabs API

   • Get an API key from ElevenLabs
   • Note your preferred voice ID

Installation

npm install

Configuration

1. Copy .env.example to .env:

   cp .env.example .env
   

2. Fill in your credentials:

   # Discord Bot Configuration
   DISCORD_BOT_TOKEN=your_discord_bot_token_here
   DISCORD_CLIENT_ID=your_discord_client_id_here
   
   # Google Gemini API Configuration
   GEMINI_API_KEY=your_gemini_api_key_here
   GEMINI_MODEL=gemini-1.5-flash-002
   
   # ElevenLabs Configuration  
   ELEVENLABS_API_KEY=your_elevenlabs_api_key_here
   ELEVENLABS_VOICE_ID=Au8OOcCmvsCaQpmULvvQ
   

Usage

Development Mode

npm run dev

Production Mode

npm run build
npm start

SSE Mode (for HTTP transport)

npm run dev:sse    # Development
npm run start:sse  # Production

MCP Tools

Bot Management

• bot_connect - Connect the Discord bot
• bot_disconnect - Disconnect the Discord bot
• bot_status - Get bot status and connection info

Voice Channel

• join_voice_channel - Join a voice channel
• leave_voice_channel - Leave the current voice channel
• speak_in_voice - Use TTS to speak in voice channel
• start_transcription - Start transcribing voice
• stop_transcription - Stop transcribing voice

Text Channel

• send_message - Send a message to a text channel

Guild Management

• list_guilds - List all guilds the bot is in
• list_channels - List all channels in a guild

Voice Transcription

The bot uses:

• Voice Activity Detection (VAD) to detect when users are speaking
• Google Gemini for accurate speech-to-text transcription
• Automatic silence detection to segment speech

Transcripts are saved to the ./transcripts directory in JSON format.

Text-to-Speech

The bot can speak in voice channels using:

• ElevenLabs API for natural-sounding speech
• Configurable voice selection
• Streaming audio playback

Architecture

discord-bot-mcp/
├── src/
│   ├── index.ts          # MCP server entry point
│   ├── config.ts         # Configuration management
│   ├── services/
│   │   ├── discord-bot.ts    # Discord bot service
│   │   ├── transcription.ts  # Voice transcription service
│   │   └── elevenlabs.ts     # TTS service
│   └── utils/
│       ├── logger.ts          # Logging utility
│       └── transcript-logger.ts # Transcript file management

Environment Variables

Testing

Quick Test with Curl (Stateless)

# Start the server
pnpm dev:sse

# Test with curl (no session needed)
curl -X POST http://localhost:3003/message \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "bot_status",
      "arguments": {}
    },
    "id": 1
  }'

# List available tools
curl -X POST http://localhost:3003/message \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "method": "tools/list", "params": {}, "id": 1}'

# Check server status
curl http://localhost:3003/status

Stateful Operations with Sessions

For stateful operations (like setting current server/channel), create a session:

# Create a session
SESSION_ID=$(curl -s http://localhost:3003/session | python3 -c "import sys, json; print(json.load(sys.stdin)['sessionId'])")

# Use the session for stateful operations
curl -X POST "http://localhost:3003/message?sessionId=$SESSION_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "view_server",
      "arguments": {"server_name": "RustyButter"}
    },
    "id": 1
  }'

# Now list channels will show channels from the current server
curl -X POST "http://localhost:3003/message?sessionId=$SESSION_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "list_channels",
      "arguments": {}
    },
    "id": 2
  }'

Sessions are maintained server-side for 30 minutes of inactivity.

License

MIT