📱 Sinch MCP GPT SMS Server

An AI-native SMS assistant powered by GPT and MessageMedia — now with a live SMS report dashboard!

✅ Features

• Send and receive SMS via GPT

• View full SMS conversation history

• Auto-reply via GPT using webhook

• Contextual memory per phone number

• MCP-compliant with /meta and /function-schema

• LangChain agent integration

• 📊 Live SMS Report Dashboard (NEW!)

Related MCP server: Twilio SMS Server

📊 Live Dashboard Preview

Below is a screenshot of the embedded high bar chart showing SMS volume:

🧠 Architecture Overview

User ↔ chat.html
     ↕
GPT Chat Server (OpenAI-router.js)
     ↕ (tool calls)
MCP Server (index.js)
     ↕
MessageMedia SMS API

• index.js: Core MCP server, webhook processor, /context + /send + /report endpoints

• Openai-router.js: ChatGPT proxy handling tool calls via OpenAI Functions

• chat.html: Interactive frontend for messaging and tools

• sms_report_live_dashboard.html: Live chart using /report

🛠️ How to Run

npm install
OPENAI_API_KEY=your_key node start-all.js

• Access Chat UI: http://localhost:3000/chat.html

• Access Report Dashboard: http://localhost:3000/sms_report_live_dashboard.html

📂 Folder Structure

.
├── index.js                # MCP Server
├── Openai-router.js        # GPT Chat Interface
├── webhook-log.json        # Legacy logs (if not using MessageMedia)
├── chat.html               # GPT chat client
├── sms_report_live_dashboard.html  # Dashboard with Chart.js
├── langchain_mcp_agent.py  # LangChain integration script
├── .env                    # Environment variables
├── public/
│   └── sms_report_preview.png  # Chart screenshot

📦 Tools

• get_sms_context

• send_sms

• show_sms_report (NEW)

📍 Roadmap

• 🔄 Auto-classification of replies

• 📈 Sentiment + urgency scoring

• 🤖 Unified Copilot with webhook/CLI/chat

📊 SMS Report Chart

A new feature has been added to display a live dashboard of SMS volume by day.

You can now ask in the chat.html interface:

"Show me my report history of +61412345678"

The assistant will respond with an embedded report like this: