Zomato MCP Server: ChatGPT & Claude can now order food for you (Step by step Guide)

MCP follows a client-server architecture with four key components:

• 1. MCP Host (The AI Application): This is your Claude Desktop, ChatGPT, or any AI-powered tool that wants to access external services.


• 2. MCP Client (The Connector): Built into the host application, this maintains connections with MCP servers and translates between the AI and the protocol.


• 3. MCP Server (The Bridge to Real Services): Each server exposes specific capabilities—like Zomato for food ordering, GitHub for code repositories, or Google Drive for documents. Servers are lightweight programs that connect to either local data (files on your computer) or remote services (APIs over the internet).


• 4. Transport Layer (The Communication Channel): MCP supports two primary methods:

STDIO (Standard Input/Output): For local integrations where the server runs on your machine HTTP + SSE (Server-Sent Events): For remote connections, enabling real-time streaming of data

All communication uses JSON-RPC 2.0 as the message standard, providing structure for requests and responses.

This entire architecture reflects how AI Development Services are evolving to connect digital systems with human-like interactivity.

Zomato open-sourced its MCP Server Manifest, showing how AI can directly interact with its ordering ecosystem, a prime example of AI integration services in real-world action.

Here’s what it enables:

Let me walk you through the complete setup process for both Claude Desktop and VS Code.

Prerequisites

Before starting, ensure you have:

• Node.js installed on your system (check by running node -v in terminal)


• A Zomato account (the same one you use on the app)


• Claude Desktop or VS Code with Claude extension

Important OAuth Limitation: Currently, Zomato has only whitelisted localhost and chatgpt.com as OAuth redirect URIs. If you're using a different environment, you'll need to reach out to Zomato to enable your client.

No restaurants found

Cause: Imprecise location data or searching outside Zomato's delivery zones

Fix: Provide specific addresses like "Koramangala 6th Block, Bangalore" instead of just "Koramangala Area"

Authentication loop (repeated OAuth prompts)

Cause: Token storage issues or expired credentials

Fix: Completely quit Claude, delete stored config at ~/Library/Application Support/Claude/, restart, and re-authenticate

Tools not appearing in Claude:

Cause: Configuration file syntax errors

Fix: Validate your JSON using a JSON validator tool. Common errors include missing commas or mismatched brackets

Connection timeout errors

Cause: Network issues or server overload

Fix: Check your internet connection, increase timeout limits, or wait and retry

Here’s the simplified workflow:

• MCP Server exposes Zomato’s capabilities (search, get_menu, add_to_cart, checkout).
• MCP Manifest defines what AI can access.
• AI Model (Claude, ChatGPT) reads the manifest and performs the action you request.

You say:
“Order biryani near me.”

It does:
You → AI → MCP Server → Zomato API → Restaurant → Checkout

No clicks. No app. Just conversation.

Zomato didn’t just make food ordering smarter, it demonstrated the future of AI integration services in everyday life. Because when AI can trigger real-world actions through MCP, your chat interface becomes your entire app. They built a framework for the next generation of digital interaction.

Here’s why it’s a game-changer:

• App-free interaction: Skip the UI, go straight to execution.
• Interoperable AI: Any MCP-compatible model can use it.
• From talking to doing: AI stops being passive and starts being productive.

This is what “AI-native”truly means.

Sure, it’s early and has it’s own Limitation.

Location Precision
AI sometimes struggles with vague inputs like “near MG Road.” The server needs exact latitude, longitude, and formatted addresses.

Fix: Use precise inputs — e.g., “Koramangala 5th Block” or paste coordinates from Google Maps.

Manual Payment
Automating payments is restricted as a security measure. QR codes and payment links will get generated. You’ll still need to scan the code or open the link yourself to finish payment.

There’s manual payment, limited coverage, and a tech setup required.

API Rate Limits
Zomato’s API allows only limited requests. Too many searches in a short time can trigger limits.

Tool Selection
AI Models might sometime pick the slower tool.

Be specific, say “Search South Indian restaurants using keyword search” instead of “Show me South Indian Options.”

The get_all_restaurants tool is detailed but slow. For quicker results, use get_restaurants_for_keyword.

Every big shift starts small and this one’s showing us the next evolution of digital experiences.

The Bigger Picture

Zomato’s MCP initiative is more than a cool hack.

It’s a blueprint for what’s coming across industries:

• AI becomes your interface.
• MCP servers become the new APIs.
• “App-free” interaction becomes the new normal.

Soon, your users won’t “open” apps, they’ll just ask them.

Building an MCP server lets AI to connect directly with your tools and APIs to place orders, update CRMs, or fetch data.

The outcome: Quicker workflows, happier customers, and smarter automation.

But there’s a bigger win, distribution. By integrating their services directly with AI models like ChatGPT or Claude, businesses can reach users right where they already are. Instead of pulling customers to their app, they can distribute their offerings inside AI ecosystems, making their product accessible to millions with a single integration.

In short, MCP servers don’t just boost efficiency — they help businesses scale reach, automate delivery, and win distribution in the AI-first world.

Zomato’s innovation proves what’s possible when AI and automation merge seamlessly.

At Rytsense Technologies, we help businesses design and build custom AI solutions that connect directly with your existing tech stack.