Travel and Hospitality Agent Readiness: Hotels, Airlines, and the Booking API Gap
Travel should be the most agent-ready category on the internet. It is, after all, the category that invented machine-readable inventory: GDS systems have been moving structured flight data since 1964. And yet when an AI agent tries to book a hotel in Paris, it ends up on Booking.com because the independent hotel has no agent-accessible booking endpoint. The whole industry sits behind intermediaries — and the intermediaries are the only ones scoring well on Agent Readiness.
The Paradox: Machine-First Industry, Agent-Last Infrastructure
Sabre was founded in 1960. American Airlines and IBM built it specifically so machines could query flight inventory and book seats without a human in the loop. That was 65 years ago. Hospitality followed with global distribution systems in the 1980s. Restaurants formalized OpenTable by 1998.
The entire travel category has been natively machine-first for decades. You would expect travel to be the category most ready for AI agents. The data says the opposite. Independent hotels average 28/100. Airlines average mid-30s. Restaurants average below 30. The only travel-adjacent companies scoring in Silver and Gold are the intermediaries themselves — Booking.com, Expedia, Kayak, OpenTable.
The reason is simple: the machine layer in travel was built for B2B distribution, not direct agent access. Hotels sold their inventory API to OTAs. Airlines sold theirs to GDS. Restaurants sold theirs to OpenTable. Consumers (and now their agents) were always supposed to go through a front door the industry never built.
Who Owns the Agent Relationship Today (And Why That Is Fragile)
Three classes of intermediaries currently own the agent-accessible surface for travel. Each of them is a layer the original supplier could have built — and each represents margin the supplier gives up to stay discoverable.
Booking.com and Expedia (Hotels)
Independent hotels have quietly ceded their customer relationship to OTAs. A buyer-side agent looking for a boutique hotel in Lisbon will hit Booking.com first because the hotel itself has no structured availability API. The OTA takes 15-25% commission and owns the agent relationship.
Score impact: Hotels score 18-42 on Agent Readiness. OTAs score 70+ because they built the API layer the hotels should have built.
GDS and NDC (Airlines)
Airlines have existed in a machine-readable world since the 1960s through GDS systems like Sabre and Amadeus. NDC is the modern successor. But none of these surfaces are agent-accessible without a commercial relationship, and no major airline publishes an MCP server or agent-card.json.
Score impact: Airlines average mid-30s on Agent Readiness. Better than hotels but still below where a naturally machine-first industry should be.
OpenTable and Resy (Restaurants)
Restaurant reservations are one of the most-requested AI agent tasks. Most restaurants route discovery through OpenTable or Resy rather than expose a direct reservation API. When an agent tries to book directly, it gets a phone number.
Score impact: Independent restaurants average 25-40. OpenTable scores in the 60s. The restaurant loses 5-10% per cover to the reservation platform.
Why this is fragile: Agents route to whichever surface exposes structured booking. Today that is the OTA. The moment a hotel chain, independent property, or restaurant exposes its own MCP server, the agent can bypass the intermediary — and most will, because it is faster, cheaper, and more reliable than screen-scraping a Booking.com results page.
The Four Tools Every Agent-Ready Travel Business Needs
AgentHermes analyzed 24 booking journeys across hotels, airlines, and restaurants. Four tools cover ~95% of what agents actually need to do.
search_availability()Query open inventory across a date range with structured filters — room type, cabin class, party size, dietary restrictions. Returns typed availability windows with prices.
get_rates()Return structured pricing including base rate, taxes, resort fees, cancellation policy, and any loyalty program discounts. Agents compare rates across properties in seconds.
create_booking()Structured booking tool with guest details, payment token, and confirmation. Returns a booking reference the agent can share with the user and use for future modifications.
get_policies()Expose cancellation, change, and refund policies as structured data. Agents need to understand liability before they commit on behalf of a user. Unstructured policy pages cause booking refusals.
A fifth tool, modify_booking(), handles the long-tail scenarios: change dates, cancel, upgrade, add a companion. It is worth shipping but will not block an agent from completing the initial transaction if missing.
The Agent-Ready Travel Playbook
Five steps that will move a hotel, airline, or restaurant from Bronze to Gold. Order matters: steps 1-2 move you to Silver, steps 3-5 move you to Gold.
Publish direct availability — even a narrow window
Start with a single endpoint: GET /api/availability?start=2026-05-01&end=2026-05-07. Return JSON with open dates and starting prices. This alone lifts your D2 API score from single digits to 70+ and moves you into Silver tier.
Structure your policies with JSON-LD
Add schema.org LodgingBusiness, Restaurant, or AirlineFlight markup with cancellationPolicy, paymentAccepted, and openingHoursSpecification. Agents read this instantly. Humans never see it. Zero trade-off.
Expose structured booking confirmation
When an agent creates a booking, return a typed response with booking_id, confirmation_code, guest details, charges, and next-step URLs. Do not return an HTML success page — agents cannot verify the booking from HTML.
Add an MCP server with your booking flow
The AgentHermes hospitality vertical template ships search_availability, get_rates, create_booking, get_policies, and modify_booking tools out of the box. No code — fill out the wizard and the MCP endpoint goes live.
Publish agent-card.json and llms.txt
Zero of the 500 businesses in our scan have an agent card. Travel businesses that ship one first become the canonical agent-accessible option in their category — especially for boutique hotels and independent airlines competing against OTAs.
What Happens When Agents Can Book Direct
OTA commissions become optional
The moment an independent hotel ships a direct MCP server with rates that match the OTA (or slightly better), agents will route there to save the user 5-15%. The OTA commission was always a tax on invisibility.
Loyalty programs compound
Direct agent bookings apply loyalty perks, status upgrades, and point redemption. OTA bookings cannot. Every direct booking is more valuable than an OTA booking even at the same rate — because the hotel captures the long-term relationship.
Restaurants escape OpenTable fees
A restaurant exposing its own reservation MCP server no longer owes per-cover fees to OpenTable or monthly fees to Resy. For a 100-seat restaurant, that is often $2-5K per month recovered.
Airlines unlock NDC for agents
NDC was built for richer content delivery through GDS. An agent-accessible layer on top of NDC lets airlines expose ancillaries (bag fees, seat upgrades, lounge access) as structured tools. More revenue per booking.
The travel industry has spent two decades watching intermediaries capture margin that should have belonged to suppliers. Agents are the one leverage point where suppliers can reclaim that relationship — but only if they move before their competitors do. Agents will not wait for the whole industry to catch up. They will route to whichever hotel, airline, or restaurant exposes the first usable API, and that behavior will calcify into a new discovery pattern.
Frequently Asked Questions
Why are hotels and airlines scoring so low when the industry is already digital?
Travel has been digital for 40+ years, but the digital layer was built for B2B distribution through GDS, OTAs, and channel managers — not direct agent access. Hotels cede their availability feed to Booking.com and Expedia in exchange for traffic. Airlines publish schedules through Sabre and Amadeus, which are gated behind commercial agreements. No one in the supply chain built a public agent-accessible API because no one needed to. Now agents exist, and the entire category is caught flat-footed.
Will AI agents actually disintermediate Booking.com and Expedia?
The OTAs will adapt — they are sophisticated companies with strong agent relationships already. What will change is which hotels win when an agent is choosing. Right now, the agent has no choice but to use the OTA because the hotel has no direct agent surface. Once a hotel ships an MCP server with direct booking, the agent can route there, skip the OTA commission, and deliver a lower price or better perk to the user. Hotels that move first capture that value. Hotels that wait stay permanently inside the OTA ecosystem.
What tools should a hotel expose in its MCP server?
The AgentHermes hospitality template exposes five tools by default: search_availability (date range + party size), get_rates (pricing with taxes and fees), create_booking (with guest details and payment token), get_policies (cancellation, check-in, amenities), and modify_booking (change dates, cancel, upgrade). Resources include property info, room types, and amenity lists. That set handles ~95% of real agent booking journeys.
Can a restaurant become agent-ready without paying OpenTable or Resy?
Yes. Restaurants can expose a thin reservation API directly — even a basic GET /api/availability and POST /api/reservation pair is enough to enter Silver tier. AgentHermes ships a restaurant vertical template with exactly these tools. The alternative is paying $200-500 per month to OpenTable or Resy for the same functionality, plus a per-cover fee. Owning the agent surface directly is cheaper and preserves the customer relationship.
How does travel agent readiness interact with existing loyalty programs?
Agent-ready travel is an unlock, not a threat, for loyalty programs. An MCP server can expose get_loyalty_status, apply_points, and redeem_reward tools. When an agent books on behalf of a Marriott Bonvoy member, it automatically applies status benefits and uses points when the user instructs it to. That experience is better than what the OTA can deliver — OTAs typically cannot apply hotel loyalty perks — which creates a direct booking incentive loop that favors properties with good agent infrastructure.
Score your hotel, airline, or restaurant in 60 seconds
See where you rank across the 9 Agent Readiness dimensions — and which booking tools, APIs, and discovery files would move you out of OTA dependency.