Food402 MCP Server

An MCP (Model Context Protocol) server that enables AI assistants to order food from TGO Yemek. Simply chat with your AI assistant to browse restaurants, build your order, and complete checkout. Works with Claude, ChatGPT (Developer Mode), and Codex CLI via MCP.

---

Clawdbot / OpenClaw (AgentSkills)

Food402 is also available as an AgentSkill on ClawHub for use with Clawdbot (OpenClaw), Claude Code, Cursor, Codex, Gemini CLI, and other AgentSkills-compatible tools.

Install via ClawHub:

clawhub install rersozlu/food402

Or manually: Copy the skill/ folder to ~/.openclaw/skills/food402/ and configure your credentials in ~/.openclaw/openclaw.json.

---

Local MCP Server (npm package)

Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "food402": {
      "command": "npx",
      "args": ["-y", "food402"],
      "env": {
        "TGO_EMAIL": "your-email@example.com",
        "TGO_PASSWORD": "your-password",
        "GOOGLE_PLACES_API_KEY": "your-google-api-key"
      }
    }
  }
}

Replace credentials with your own. GOOGLE_PLACES_API_KEY is optional (see Optional Features).

Claude Code

For project-specific installation with Claude Code:

npm install food402

This automatically adds food402 to your .mcp.json. Open the file and update your credentials:

{
  "mcpServers": {
    "food402": {
      "command": "node",
      "args": ["./node_modules/food402/dist/src/index.js"],
      "env": {
        "TGO_EMAIL": "your-email@example.com",
        "TGO_PASSWORD": "your-password",
        "GOOGLE_PLACES_API_KEY": "your-google-api-key"
      }
    }
  }
}

GOOGLE_PLACES_API_KEY is optional (see Optional Features).

Codex CLI (Terminal)

Codex reads MCP servers from your global config at ~/.codex/config.toml.

Option A: Via CLI

codex mcp add food402 --env TGO_EMAIL=your-email@example.com --env TGO_PASSWORD=your-password --env GOOGLE_PLACES_API_KEY=your-google-api-key -- npx -y food402

Option B: Manual config

Add to ~/.codex/config.toml:

[mcp_servers.food402]
command = "npx"
args = ["-y", "food402"]

[mcp_servers.food402.env]
TGO_EMAIL = "your-email@example.com"
TGO_PASSWORD = "your-password"
GOOGLE_PLACES_API_KEY = "your-google-api-key"  # Optional

---

Prerequisites

Account Setup Required

Before using this MCP server, you must have a TGO Yemek account with:

1. TGO Yemek account - Create one at tgoyemek.com if you don't have one
2. Payment card saved to your account - The checkout process requires a saved card; you cannot enter card details during ordering
3. At least one delivery address saved (recommended) - You can add addresses through the MCP, but having one pre-configured makes ordering faster

Quick Start: Ordering Flow

Here's the typical workflow when ordering food through the AI assistant:

1. Select Delivery Address

"Show me my saved addresses"
"Select my home address for delivery"

2. Find Restaurants

"What restaurants are near me?"
"Search for pizza restaurants"
"Find places that serve lahmacun"

3. Browse Menu & Add Items

"Show me the menu for [restaurant name]"
"Add 2 lahmacun to my cart"
"What customization options are available for this item?"

4. Review & Checkout

"Show me my basket"
"Remove the drink from my order"
"I'm ready to checkout"

5. Place Order

"Place my order using my saved card"

Note: A browser window will open for 3D Secure verification. Complete the verification to finalize your order.

6. Track Order

"What's the status of my order?"
"Show me my recent orders"

Available Tools

Tool	Description	Parameters
get_addresses	Get user's saved delivery addresses	None
select_address	Select delivery address (must call before ordering)	addressId
get_restaurants	Search restaurants near a location	latitude, longitude, page?
search_restaurants	Search restaurants and products by keyword	searchQuery, latitude, longitude, page?
get_restaurant_menu	Get restaurant's full menu	restaurantId, latitude, longitude
get_product_details	Get product customization options	restaurantId, productId, latitude, longitude
get_product_recommendations	Get "goes well with" suggestions	restaurantId, productIds[]
add_to_basket	Add items to cart	storeId, items[], latitude, longitude, etc.
get_basket	Get current cart contents	None
remove_from_basket	Remove item from cart	itemId
clear_basket	Clear entire cart	None
get_cities	Get list of all cities for address selection	None
get_districts	Get districts for a city	cityId
get_neighborhoods	Get neighborhoods for a district	districtId
add_address	Add a new delivery address	name, surname, phone, addressName, addressLine, cityId, districtId, neighborhoodId, latitude, longitude, etc.
get_saved_cards	Get user's saved payment cards (masked)	None
checkout_ready	Get basket ready for checkout with payment context	None
set_order_note	Set order note and delivery preferences	note?, noServiceWare?, contactlessDelivery?, dontRingBell?
place_order	Place order with 3D Secure (opens browser for verification)	cardId
get_orders	Get user's order history with status	page?
get_order_detail	Get detailed order info including delivery status	orderId

Optional Features

Google Reviews Integration

The get_google_reviews tool allows you to fetch Google Reviews for restaurants, enabling comparison between TGO ratings and Google ratings. This feature is optional and requires a Google Places API key.

To enable Google Reviews:

1. Get a Google Places API key from the Google Cloud Console
2. Enable BOTH APIs for your project (both are required):
   • Places API (New) - Used for text search to find restaurants
   • Places API - Used for fetching place details and reviews
3. Add the key to your MCP server configuration:

{
  "mcpServers": {
    "food402": {
      "command": "npx",
      "args": ["-y", "food402"],
      "env": {
        "TGO_EMAIL": "your-email@example.com",
        "TGO_PASSWORD": "your-password",
        "GOOGLE_PLACES_API_KEY": "your-google-api-key"
      }
    }
  }
}

Without the API key: The tool will return a helpful message indicating the feature is unavailable but won't cause any errors. The rest of the food ordering functionality works normally without it.

---

Development

Repository Structure

food402/
├── src/                    # MCP server (stdio transport)
│   ├── index.ts            # MCP entry point with tool definitions
│   ├── auth.ts             # TGO auth with token caching
│   ├── api.ts              # Thin wrapper around shared/api.ts
│   └── postinstall.ts      # Auto-configures .mcp.json on npm install
├── shared/                 # Shared API code
│   ├── api.ts              # Token-parameterized TGO API functions
│   └── types.ts            # TypeScript interfaces
├── test/                   # Test suite
│   ├── integration/        # API integration tests
│   ├── security/           # Security validation tests
│   └── setup/              # Test utilities and fixtures
├── package.json            # Root package (npm: food402)
├── README.md
└── CLAUDE.md

Local Server Development

# Install dependencies
npm install

# Run in development mode
npm start

# Build TypeScript
npm run build

Testing

The project includes integration tests and security tests using Vitest.

# Run all tests
npm test

# Run security tests only
npm run test:security

# Run tests with coverage
npm run test:coverage

Setup for tests:

1. Copy test/.env.test.example to test/.env.test
2. Add your TGO credentials to test/.env.test
3. Run npm test

Tests cover:

• Address management (cities, districts, neighborhoods)
• Restaurant discovery and menu browsing
• Basket/cart operations
• Payment card retrieval and checkout readiness
• Order history
• Security checks for credential handling

License

MIT