Mood-Based Music & Quote MCP Server

An MCP-style server that reacts to WhatsApp/SMS messages: it classifies the user's mood, fetches a curated Spotify playlist, adds a motivational/funny quote, and can optionally deliver a voice note (ElevenLabs). Includes Twilio webhook for inbound WhatsApp and simple MCP endpoints for tool calls.

Quick start

1. Install deps

npm install

2. Copy env template

cp config/env.example .env

Fill in Spotify client credentials, Twilio (for WhatsApp/SMS), and optional ElevenLabs + OpenAI.

3. Run dev server

npm run dev

Endpoints

• GET /healthz – health check

• GET /mcp/tools – list available tools

• POST /mcp/call – invoke a tool: { "name": "run_mood_flow", "input": { "to": "whatsapp:+1...", "text": "I feel stressed", "includeVoice": false } }

• POST /webhooks/twilio – Twilio WhatsApp/SMS webhook; echoes back playlist + quote

Tools exposed

• detect_mood

• get_playlist

• get_quote

• synthesize_voice (if ElevenLabs configured)

• send_message

• run_mood_flow (end-to-end orchestration)

Notes

• Voice replies require hosting the MP3 somewhere accessible; the current stub generates audio but does not upload it. Add an object-store upload to send MediaUrl with Twilio.

• Mood classification defaults to a keyword heuristic; plug in an LLM if desired.

• MCP shape here is HTTP-based for simplicity; can be adapted to an MCP transport.