UCP Lodging makes hotels bookable by AI agents.

What was released

On September 25, the Universal Commerce Protocol added booking as its first lodging capability, dev.ucp.lodging.booking. It was shaped by a Lodging Tech Council that includes Amadeus, Booking.com, Expedia, Google, Hilton, Marriott, and Trip.com, and it is now open for public feedback.

The spec gives any AI surface a consistent way to run a hotel booking session. An agent creates a session from a property, room type, and rate plan, adds guests and occupancy, receives real-time pricing and availability, and completes payment. It handles deposits, deferred balances, pay-at-property, and split payments. Price changes mid-session are displayed to the guest, and taxes and fees are shown up front. The hotel stays the merchant of record. When a step needs the hotel's own experience, the session hands back to the hotel's booking flow.

This is a big step for hotels. A shared standard backed by leading hotel groups, distribution partners, and Google means less custom integration work, a guest experience built on transparent pricing, and a clear route for agents to book directly with the property. We're excited to see the industry build it together.

A strong foundation, with room to grow

This first release focuses on the transaction. The spec describes discovery as "provisional" and booking as "authoritative." UCP's broader search and catalog capabilities already work at the hotel level, and we expect the lodging work to extend upstream from here as the community builds on it.

In the draft, the property object carries an ID, a name, an address, and media. Room types and rate plans carry a title and a free-text description. The richer context that helps an agent match a traveler to the right property lives alongside the protocol today, and that's where the next wave of work will happen.

Search is changing, and hotels are well placed

An agent booking a hotel does two jobs. First it helps the traveler choose a property. Then it books it. UCP now gives the industry a common answer for the second job. The first is changing just as fast, as travelers ask AI to shop, compare, and decide on their behalf.

Once agents can book, the quality of what AI knows about a hotel carries straight through to the reservation.

The spec already points this way. Rate plan titles and descriptions have to convey the material commercial terms on their own and agree with the policies attached to the booking. That rewards brands whose content, policies, and offers tell one consistent story everywhere AI encounters them. A hotel that AI understands accurately, with a bookable endpoint, is ready to be recommended and booked on its own terms.

Where Bonafide fits

Bonafide is additive to UCP. We operate the Context Layer that informs the booking flow. It turns fragmented enterprise truth, from property content and policies to offers and loyalty, into verified, current context that AI can use to recommend a property for the right reasons and route the guest to the right booking path.

Open distribution has been part of our architecture from the start, with APIs, MCP, and UCP as delivery channels. As UCP grows upstream, the same context that improves how AI describes your properties is what keeps an agent's recommendation consistent with what the booking session returns. We're reviewing the draft closely and look forward to contributing to the UCP community alongside Google and the council.

Questions to start with