restaurant-mcp

MCP server for Spice Garden, a demo restaurant, with an optional OpenRouter-powered chat client. The assistant queries a live SQLite database for orders, inventory, tables, and sales data.

Prerequisites

• Node.js 20.6+ (required for --env-file)

• An OpenRouter API key (for the chat client only)

Related MCP server: mcp-chinookdb-server

Quick start

npm install
cp .env.example .env   # Windows: copy .env.example .env
# Edit .env — add your OPENROUTER_API_KEY and pick a tool-capable model
npm run seed
npm run chat

Example questions:

• How many orders today?

• Which is our most popular dish?

• What's running low in inventory?

• How many tables are occupied?

Type exit to quit.

Scripts

Environment variables

Copy .env.example to .env:

Choosing a model

The chat client sends tools with every request, so the model must support function/tool calling. Browse compatible models at:

https://openrouter.ai/models?supported_parameters=tools

Free-tier examples that work:

• meta-llama/llama-3.3-70b-instruct:free

• qwen/qwen3-coder:free

• openai/gpt-oss-20b:free

MCP server tools

Use with Cursor (or any MCP client)

Add to your MCP config:

{
  "mcpServers": {
    "restaurant": {
      "command": "node",
      "args": ["/absolute/path/to/restaurant-mcp/server.js"]
    }
  }
}

Run npm run seed first. The server exits with a clear message if the database is missing or empty.

Project structure

restaurant-mcp/
├── server.js       # MCP server (stdio)
├── client.js       # OpenRouter chat client
├── db/
│   ├── seed.js     # Database schema + sample data
│   └── restaurant.db  # Generated (gitignored)
├── lib/
│   ├── db.js       # Database bootstrap check
│   └── paths.js    # Shared paths and version
└── .env            # Your secrets (gitignored)

License

ISC