Product · AI agents
Restaurant commerce that AI assistants can use
This page is available in English for now.
People are starting to ask AI assistants where to eat and what to order. Thmenu does not try to be the assistant; it gives assistants a reliable way to work with restaurants. Through Thmenu's MCP server an assistant can find a listed restaurant, read its live menu, check reservation slots, request a booking and prepare an order — and the guest confirms before anything is placed or paid.
Discovery → Availability → Reservation → Menu → Order
Discovery
The assistant searches restaurants that chose to be listed in the Thmenu directory.
Menu and availability
It reads the live menu and the open reservation slots — the same data your guests see.
Reservation or order
It can request a reservation, or build a cart and, with the guest's sign-in, place an order.
The guest confirms
Payment is never taken by the assistant; the guest finishes on the restaurant's own confirmation page.
How it is built
- 01
An open protocol
Thmenu speaks the Model Context Protocol (MCP), which assistants use to call tools.
- 02
Proper sign-in
Actions on a guest's behalf use OAuth 2.0 with PKCE and narrow scopes, so an assistant can only do what the guest allowed.
- 03
Restaurant opt-in
A restaurant appears to assistants only if it chose to be listed, and accepts assistant orders only if it switched them on.
- 04
Orders are labelled
Orders that came through an assistant are marked as such in the staff orders screen.
Good to know
- Assistants differ
- Whether a given assistant can use MCP tools depends on that assistant. The menu, reservation and order data stay the same either way.
AI agents