ADAPT

Every AI agent. Direct to your hotel.

MCP servers, CLI tools, REST APIs, and programmable settlement — the open protocols that let AI agents discover, book, and pay your property directly. No OTA. No 15–25% commission. You own the guest data.

Read the manifesto
  • Live pilot · The Exchange Building, Memphis
  • Open alliance · Formed 2025 · UNA 2026
Master display · protocol programPilot · Memphis · in production
claude — MCP hotel bookingSample exchange · demo datalink
0510152025
<8%Commission OTA band · 15–25 %
Programmable settlement · on

Terms enforced by protocol. Payment clears instantly; cancellation and settlement conditions travel with the booking.

The OTA tax on every booking

For 25 years, OTAs have extracted 15–25% of every hotel booking. Now AI agents are the new distribution channel — and most properties are invisible to them.

Commission extracted per booking by Booking.com & Expedia15–25 %
Annual OTA commissions extracted from hotels globally$50B+
Guest email & phone — withheld; hotels collect manually at check-inname only
Extranet corrections, bulk edits, reconciliationobstructed

The old way

Every booking passes through 5–7 gatekeepers, each taking a cut

Traveler

Searches on

GoogleTripAdvisorKayak

Metasearch−CPC fees

Booking.comExpediaAirbnb

OTAs−15–25 % commission

Channel MgrGDS Switch

2-way XML/SOAP sync (SiteMinder, DHISCO)−SaaS fees

CRSPMSRMS

Opera, Mews, Cloudbeds−per-room/mo licensing

VCC PaymentsMerchant

Virtual credit cards, reconciliation−processing fees

Hotel keeps ~75 %

Deceptive design in OTA extranets

OTA operator portals employ dark patterns — deliberate UI/UX choices that make it harder for hotels to reduce commissions owed. Early departures, partial refunds issued weeks after checkout, and rate adjustments are buried in multi-step workflows designed to discourage corrections. Bulk editing multiple reservations is either unavailable or deliberately cumbersome. Monthly statements arrive in formats that resist easy reconciliation.

The result: night auditors and revenue managers — often the most junior staff handling these adjustments — give up on disputed line items because the time cost of correcting a $30 commission exceeds the recovery. Across thousands of properties, these small concessions compound into billions in overpaid commissions annually. The interface isn't broken. It's working exactly as designed.

5–7 gatekeepers3–5 separate systems2-way XML/SOAPRate parity clausesGuest data lost

The ADAPT way

Direct connection. One platform. Under 8% commission.

TravelerAI Agent

CLI, ChatGPT, Claude, Gemini, Perplexity, Browser

MCPCLIRESTMarkdown

Open protocols — JSON, no XML, no switch

Hotel keeps 92 %+

Under 8 % commission

1 platformDirect connectionJSON/MCPProgrammable settlement100% guest data

AI agent discovers. Books. Done.

Three exchanges between a traveler's agent and your property. No extranet in between.

  1. Traveler asks an AI agent

    "Find me a boutique hotel in Memphis for next weekend" — via Claude, ChatGPT, Gemini, Perplexity, or any AI assistant.

  2. Agent discovers your property via MCP

    Your ADAPT MCP server responds with real-time availability, rates, room types, and photos. The AI agent evaluates options for the traveler.

  3. Guest books direct — you own the relationship

    Booking created through your MCP server. Programmable settlement clears payment instantly — terms enforced by protocol, not by trust. Under 8% commission. You retain 100% of guest data.

Four ways AI agents find you

ADAPT publishes your property across every protocol AI agents use to discover and book inventory. All four channels carry the same inventory.

CH 1 published

MCP Server

Model Context Protocol server exposing search_availability(), getRates(), and createBooking() tools. Any MCP-compatible AI agent can discover and transact directly.

TransportJSON-RPC over HTTP
Toolssearch · rates · book
ClientsClaude · ChatGPT · Gemini
server.tool("search_availability", { city, dates })
CH 2 published

CLI Tools

Terminal-first hotel search and booking. Install the building-os CLI to search, check rates, and book rooms from any command line or agent framework.

Binarybuilding-os
Verbssearch · rates · book
Runs inany shell or agent
building-os search --city memphis --dates 03/15-18
CH 3 published

REST API

Standard JSON/OAuth 2.0 endpoints. Auto-generated SDKs via OpenAPI spec. Works with any language, any platform, any automation pipeline.

FormatJSON
AuthOAuth 2.0
SDKsgenerated from OpenAPI
GET /api/v1/availability?city=memphis&checkin=...
CH 4 published

Agentic Markdown

Zero-infrastructure discoverability. Publish a .well-known/hotel.md file with structured availability — like robots.txt but for hotel inventory.

File.well-known/hotel.md
Infranone — a static file
Analogyrobots.txt for rooms
/.well-known/hotel.md → llms.txt for hotels

Sixty years of hotel distribution, one row.

From SABRE's first mainframe in 1960 to Priceline's $133M acquisition that became a $170B company — to MCP servers making OTAs optional. The complete interactive history with 60+ events, system architecture diagrams, and M&A ownership maps.

Read the full deep dive
The Exchange Building lit at night against the Downtown Memphis skyline 9 N. Second Street · Memphis, TN

Running in production.

ADAPT is the open alliance building the protocols that connect properties directly to AI agents. MCP servers, CLI tools, REST APIs — all publishing the Exchange Building's inventory to the agentic web. Founded by owner-operators who needed it.

200+Units
19Floors
<8%Commission
100%Data owned
Read the manifestoJoin a working group