IndiaRestros MCP server for Claude and Cursor: find Indian restaurants and dishes from an AI agent (2026)
IndiaRestros MCP server for Claude and Cursor: find Indian restaurants and dishes from an AI agent (2026)
Updated 12 November 2026 (IST). By Sunny Mittal, marketing lead at IndiaRestros.
Disclosure: I work on marketing at IndiaRestros (indiarestros.com). This guide explains how to connect our MCP server and compares it fairly with other ways an AI agent can look up restaurants in India. Every number below comes from the server’s own about_dataset tool, called on the day of writing.
The short answer
Add https://indiarestros.com/mcp as a remote (Streamable HTTP) MCP server in Claude, Cursor, VS Code, ChatGPT developer mode, Gemini CLI or Codex. You don’t need an account or an API key for the public tools. Then ask things like “veg restaurants open now in Indiranagar” or “masala dosa under ₹200 in Bengaluru”. Full set-up steps are on the IndiaRestros MCP page.
What the server covers
When I called about_dataset on 12 November 2026, it reported:
- 1,703,804 restaurants across 9,827 cities, 18,168 localities and 36 states and union territories
- 1,342,005 of them with a phone number
- Biggest cities: Bengaluru (117,295), Delhi (75,625), Hyderabad (68,497), Chennai (53,760), Mumbai (53,349) and Pune (42,090)
- Kinds of place: restaurants (801,920), cafes (237,623), fast food (218,172), bakeries (153,711), sweet shops (94,967), cloud kitchens (68,303), dhabas (51,539), food courts (41,320) and bars (36,249)
Sources and ranking rules are published at indiarestros.com/sources and indiarestros.com/ranking.
Set-up in one step
| Client | How to connect |
|---|---|
| Claude (web, desktop, mobile) | Settings → Connectors → Add custom connector, URL https://indiarestros.com/mcp, no sign-in |
| Claude Code | claude mcp add --transport http indiarestros https://indiarestros.com/mcp |
| Cursor / Windsurf / others | {"mcpServers": {"indiarestros": {"url": "https://indiarestros.com/mcp"}}} (Windsurf calls the key serverUrl) |
| VS Code | code --add-mcp '{"name":"indiarestros","type":"http","url":"https://indiarestros.com/mcp"}' |
| Gemini CLI | gemini mcp add --transport http indiarestros https://indiarestros.com/mcp |
| Codex | [mcp_servers.indiarestros] with url = "https://indiarestros.com/mcp" in ~/.codex/config.toml |
No MCP client? Plain HTTPS works too, for example https://indiarestros.com/api/v1/restaurants?q=biryani+in+kondapur. The REST reference is at indiarestros.com/api/docs.
The tools
| Tool | What it returns |
|---|---|
search_restaurants |
Places by words and location (city, locality, or lat/lon and radius), kind, cuisine and diet, with open_now in IST, cost for two and price band |
search_dishes |
Menu items across India: dish name, price in INR, veg mark, menu section and the restaurant |
get_restaurant |
Full record: address, geo, phones, website, raw and parsed timings, ratings, sources and confidence |
get_menu |
Menu sections and items with INR prices and veg marks, with the menu source and date |
list_places / get_place |
States, cities and areas with counts by kind and cuisine |
report_correction |
Sends a fix (dead phone, new timings, closure); the agent should ask the user first |
about_dataset |
Counts, sources and how ranking works |
A real example
I asked search_dishes for “masala dosa” in Indiranagar, Bengaluru. The first results were Rava Masala Dosa at ₹185, Podi Masala Dosa at ₹169 and Ghee Podi Masala Dosa at ₹184 from one Indiranagar restaurant. Each result was marked veg and linked to the restaurant’s IndiaRestros page. The tool adds a note that prices come from menus (websites, PDFs, photos) and may have changed, so an agent should tell users to confirm with the restaurant.
How it compares with other ways to give an agent restaurant data
| Option | India coverage | Menu prices | Key or account needed | Notes |
|---|---|---|---|---|
| IndiaRestros MCP | 1.7M places, 9,827 cities | Yes, where a menu is on file | No (public tools) | Built for India; open_now in IST |
| Google Maps Platform (Places API) | Broad | No | API key and billing account | Strong on reviews and photos; paid per request |
| Foursquare Places API | Broad, urban-heavy | No | API key | Good global place data |
| OpenStreetMap (Overpass API) | Varies by city | No | No | Open data; quality depends on local mappers |
| Delivery apps (Zomato, Swiggy) | Broad in metros | In-app only | Not open to third-party agents | Best for actually ordering |
When to choose something else: if you need global coverage outside India, user reviews and photos at scale, or the ability to place an order, a global places API or a delivery app is the better tool. If your agent needs India-wide dish-level prices, veg filters and Indian timings without a key, start with IndiaRestros.
FAQ
Is it free? The public tools need no account and no key.
Does it work in ChatGPT? Yes, through developer mode. Create an app with the URL and choose no authentication.
Can the agent change data? Only through report_correction, which is checked before anything goes live. Ask the user before sending one.
Disclosure: I work on marketing at IndiaRestros (indiarestros.com).
Write a comment