MCP Demo — SQLite + Live Weather

A practical MCP demo where GPT-4o can query a real database and fetch live weather — two completely unrelated capabilities — through a single MCP server.

mcp-demo/
├── server.py        # MCP server: SQLite tools + weather tool
├── seed.py          # Populates ecommerce.db with sample data
├── client.py        # OpenAI GPT-4o agentic loop
├── requirements.txt
└── README.md

---

Why this showcases MCP's value

The server owns two completely unrelated capabilities:

Tool group	Tools	Backed by
SQLite	list_tables, query_db, execute_db	ecommerce.db — local SQLite
Weather	get_weather	wttr.in public API

The client (client.py) discovers both at startup — zero hardcoded tool logic. Swap GPT-4o for Claude Desktop or Cursor tomorrow; server.py doesn't change at all.

---

Setup

python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
export OPENAI_API_KEY="sk-..."                       # Windows: set OPENAI_API_KEY=...

# Seed the database (run once)
python seed.py

---

Run

python client.py

client.py spawns server.py automatically — no separate terminal needed.

---

Try these queries

# Pure DB
What are our top 3 customers by total spend?

# DB + filter
Show all pending orders with customer names and product names.

# Cross-tool: DB to find a city, weather to check it
Who is our highest-spending customer and what's the weather like in their city right now?

# Write operation
Mark order 5 as shipped.

# Follow-up read
Show all orders for Carlos Mendes now.

---

Database schema

customers (id, name, email, city, joined_at)
products  (id, name, category, price)
orders    (id, customer_id, product_id, quantity, status, ordered_at)

10 customers across 10 cities worldwide · 10 products · 20 orders.

---

Key dependencies

Package	Purpose
mcp	Official Python MCP SDK (server + client)
openai	GPT-4o via OpenAI API
requests	HTTP calls to wttr.in weather API