GDS Integration Explained: Amadeus vs Sabre vs Travelport

Arbisoft Editorial TeamPosted on September 8, 2026

14-15 Min Read TimeAdd as preferred on Google

You've built the product. Search works, checkout works. And there's nothing to sell, because there's no inventory in it.

Every travel build hits this wall. GDS integration for a travel website is where most teams stall. We've been writing travel software since 2007 and have been KAYAK's engineering partner for over 18 years. The inventory problem stops more projects than any technical challenge we see.

The standard advice for getting past it: sign up for the Amadeus self-service developer portal, grab a key, make your first call. That advice held for years. It stopped working in July 2026, and most of the internet hasn't noticed.

What is GDS integration?

GDS integration means connecting your travel website to a global distribution system (Amadeus, Sabre, or Travelport) so it can search and book live flight and hotel inventory. The contract comes before the code. All three require a signed agreement, vetting, and certification before you touch live inventory.

Amadeus closed its free self-service tier in July 2026, so every remaining route there runs through sales. Direct GDS access typically takes six to fourteen months. An aggregator like Duffel or Kiwi.com takes one to four months and needs no travel-agent accreditation.

If you read one section, read the three gates.

What changed in July 2026: the Amadeus Self-Service shutdown

Is the Amadeus Self-Service API still available? No. The self-service section of the Amadeus for Developers portal was decommissioned on 17 July 2026 and self-service API keys were deactivated.

The sourcing is mixed, and worth stating precisely. The date comes from a letter Amadeus sent to users, reported by PhocusWire in February 2026, with a spokesperson stating that only the self-service section was being decommissioned. Amadeus's own public statement never mentions the shutdown or the date; it responds to the coverage by clarifying that its Enterprise Portal and Enterprise APIs remain fully available. The corroboration sits in public code: the amadeus4dev GitHub organisation was archived on 17 July 2026, and the Python, Node and Java SDKs are now read-only behind a deprecation notice.

Is there still a way in for a small team? Yes, though nothing self-serve. Amadeus points to its Ventures and Partners programmes and runs an Amadeus for Startups track, which screens applicants (4,000+ screened, 40+ active) and offers a contact-sales form. Even the startup door at Amadeus opens onto a sales conversation.

The lesson holds even if you never touched Amadeus. This category is contract-gated end to end.

Where do travel sites actually get their flight prices?

A global distribution system (GDS) is a commercial intermediary. It collects flights, hotel rooms and cars from thousands of suppliers, then resells access through APIs and agent terminals. When a booking site shows fares from twelve airlines at once, a GDS (or an aggregator sitting on top of one) is why.

A GDS is not a developer API product. You can't sign up and start selling the way you can with Stripe. "Content" is the industry word for the fares and inventory you're permitted to see, and your contract defines it. Two companies can integrate the same GDS and see different inventory. Treat it as a wholesale supply agreement that happens to be delivered over HTTP.

One thing you can stop worrying about: you will not need to learn a green screen. GDS terminal training takes months of supervised practice, but that's for travel agents booking by hand. You're integrating an API into a website.

Before you choose a GDS, pass three gates

Most guides open by counting airlines and hotels. That's the last thing that should drive your decision, because two earlier questions can rule out direct GDS access entirely.

Gate 1: can you legally issue a ticket?

To search fares, no accreditation is needed. To issue a ticket, you need accreditation from IATA (or ARC, the Airlines Reporting Corporation, in the United States), or you need somebody else's ticketing authority. Two different permissions. Conflating them is expensive, because teams discover it late, with a built product and a launch date.

Three ways through:

• Your own IATA or ARC accreditation. IATA's published application fees are 250, 1,500 or 2,500 CHF depending on tier, with annual fees banded by your sales through IATA's Billing and Settlement Plan (BSP). IATA's brochure states these were valid as of February 2025, main entity only, so check current rates for your market. IATA also says its GoLite tier requires no minimum financial guarantee to issue tickets, so the widespread five-figure-guarantee claim is wrong for at least one tier.
• A consolidator or host agency. An accredited business that lets you issue tickets under its credentials. Cheaper and faster, with less margin control.
• An aggregator's managed accreditation. Duffel states that with Managed Content you can sell flights without being IATA or ARC accredited, using their accreditations and airline agreements. Published cost: 1% of order value on top of the per-order fee.

Gate 2: can you forecast volume twelve months out?

GDS economics are fixed-cost-forward. Setup fee, monthly minimums, certification, all payable before your first booking. Aggregator economics are purely variable.

If you can't forecast volume, a GDS contract is a bet on demand you haven't proven. Don't take that bet pre-revenue.

The arithmetic, using Duffel's published fees. On a $400 ticket with managed content, a booking costs about $7.00 ($3.00 per order plus 1% of order value). Against a $1,000 monthly GDS minimum, that's roughly 143 bookings a month before the minimum wins on variable cost alone. Against $2,500, roughly 357.

Treat those as a floor. The GDS side also carries setup, certification, per-segment fees on top of the minimum, accreditation, and six to fourteen months before your first sale. Pushing the other way, your own accreditation kills the 1% content fee and gives you margin control an aggregator won't. Our read: below a few hundred bookings a month the arithmetic doesn't favour direct GDS, and the gap widens once you price in time.

Gate 3: what do you actually need to sell?

Now coverage matters. Direct GDS access wins when you need corporate and chain content, negotiated fares you've arranged yourself, or deep full-service-carrier inventory in one region. It matters less for leisure platforms, low-cost-carrier routes and independent hotels.

Do you need more than one GDS? Rarely at launch. Most run one and supplement it.

Amadeus vs Sabre vs Travelport: what actually differs

Airline counts are undated marketing figures that move constantly. Four things shape the choice between Amadeus, Sabre and Travelport.

The 2026 story is the divergence. Amadeus closed its open door. Asked directly whether its own developer access would be restricted, Sabre answered no, calling open developer access fundamental to its platform strategy rather than a temporary programme, and noting that while some platforms are closing self-service portals, it's expanding. That's Sabre's marketing position, so weigh it accordingly. Observers did independently note that Sabre has been adding agentic AI features to its developer portal.

Travelport sits between the two, and the detail matters. It offers a 30-day sandbox, but its own documentation states that sandbox credentials are for evaluation purposes only and that you must not conduct development without signing a contract. Pre-production credentials arrive after the contract, then certification follows. So Travelport lets you look. Sabre lets you build. In 2026 that's a live selection criterion, and we haven't seen another comparison treat it as one.

Conditionally: European or Middle Eastern market with volume and accreditation, Amadeus. Prototyping before you commit, or North American corporate focus, Sabre. Hotel-heavy with an existing relationship, Travelport. Pre-revenue, none of them yet.

What GDS access actually costs

Amadeus, Sabre and Travelport all decline to publish pricing. Each wants you to contact them, and the quote depends on your market, your forecast volume, and how much content you want. Sabre confirms it doesn't publish fixed pricing. These are genuinely negotiated contracts, which is why every figure you find online is an estimate, ours included.

Verified published pricing. Duffel's public pricing page, checked 3 September 2026:

Estimates for direct GDS. No rate card exists. The ranges below come from travel-technology integrators publishing indicative figures. Read them as an order of magnitude:

That excess-search line catches people out. Metasearch products generate enormous search-to-book ratios by design. KAYAK, whose platform we help build, processes billions of searches a year. If your product is search-heavy and booking-light, per-search charges can become your largest line item. Model your ratio before your bookings.

Your own build cost sits outside all of this. Most articles fold engineering into "GDS cost." They're separate budgets, and engineering is usually the larger one in year one.

Verify every number yourself, ours included. Duffel's help centre currently states a 0.5% Managed Content fee while its live pricing page says 1%. If a vendor's own documentation disagrees with itself, treat third-hand blog figures accordingly.

How to integrate a GDS API, phase by phase

Writing code is phase four of six. The sequence below is broadly the same for Amadeus, Sabre and Travelport.

1. Commercial qualification. You apply, they vet you. Business details, trading history, accreditation or your host's credentials. They can say no.
2. Contract and credentials. Terms, minimums and per-segment fees get settled here, so your unit economics are decided in this phase. Bring someone who reads contracts.
3. Sandbox build. Sandbox data is not live inventory, and sandbox prices are not real fares. Build against it, but don't demo it to an investor as a working booking engine. We have watched that go badly.
4. Integration development. Error handling, caching within supplier limits, idempotency on booking, post-booking flows.
5. Certification. The GDS tests your implementation before letting you near production. It works like a driving test, and you can fail. This is the phase project plans forget.
6. Production and servicing. Cancellations, exchanges, schedule changes, disruptions. Ongoing work.

Phase six is where the real work lives, and nobody scopes it properly. When we built the airline-facing side of Travelliance's platform, the interesting engineering sat in the disruption path: letting airline systems request accommodation for large disruptions, sometimes up to 500 passengers at once, then reconcile and invoice all of it afterwards. Booking a seat is a solved problem. Everything that happens when the plan changes is still hard.

Travel API aggregators: the alternative to direct GDS access

An aggregator sits on top of multiple sources (GDS content, direct airline NDC connections, low-cost carriers) and gives you one API instead of several contracts. Duffel and Kiwi.com are the usual starting points, with TBO Air and Travelfusion filling low-cost carrier gaps.

What you gain: no upfront cost, no accreditation requirement, one to four months to launch, and costs that scale with revenue.

What you give up: a per-booking fee forever, partial coverage against a full GDS, dependence on someone else's accreditation and airline agreements, and less control over negotiated fares.

There's a real objection to our own advice, and the Amadeus Self-Service shutdown in July 2026 is it. The teams hurt worst were the ones who'd built on a tier they didn't control. Depending on any single supplier's goodwill carries risk, and an aggregator is still a single supplier.

So the mitigation matters more than the choice. Keep supplier-specific logic behind an internal interface. KAYAK aggregates hundreds of providers, and its architecture exists precisely so that no one provider's change becomes an emergency. The principle scales down. If swapping a supplier means touching one adapter instead of your whole codebase, a shutdown notice costs you a sprint instead of a quarter.

On NDC: it's a data standard for airlines to distribute richer offers, not a supplier you can call. Aggregators and GDSs both carry NDC content. You don't need a strategy for it on day one. You need to know your provider supports it.

Starting on an aggregator and migrating to direct GDS later is the normal path. It's what we recommend to most teams who ask.

Which API should you use to sell hotel inventory?

For most travel platforms, a hotel wholesaler (a bedbank) rather than a GDS. Hotel inventory follows a different supply chain from flights.

GDS hotel APIs were built on airline reservation infrastructure from the 1990s. They're strong on major chains and corporate rates, weaker on independents, which is why most OTAs pair a GDS with at least one bedbank.

A useful ceiling, reported by hotel-aggregation specialists working across 100+ live integrations: four to six well-chosen suppliers (one global bedbank, one GDS, one OTA-wholesale source, one or two regional specialists) cover the large majority of globally bookable inventory. Past six, each new integration adds maintenance and duplicate properties faster than it adds unique hotels. Those specialists sell aggregation, so read it as directional. Either way, resist the urge to collect suppliers.

Build GDS integration in-house, or hire a partner?

The hard parts of a GDS integration are fare rules, timeout and error handling, caching inside supplier limits, idempotency on booking, and servicing after the sale. Teams underestimate the last one almost universally.

The maintenance burden is real and current. Travelport deprecated all of its previous JSON API authentication endpoints on 30 January 2026 and issued new credentials to customers, with transactions to the old endpoints failing. That's a normal Tuesday in travel integration.

If you're hiring a partner, ask:

• Which suppliers have you certified with, and when? A 2019 certification isn't current knowledge.
• Who owns the supplier contract, you or them?
• What happens when a supplier deprecates an endpoint? Included, or change request?
• Does scope stop at launch, or cover servicing?

We're not going to publish a list of "top GDS integration companies," ourselves included. Those four questions will tell you more than any ranking. Our vetting checklist for travel software partners goes deeper.

Your next step

FAQs

Is the Amadeus API free? 

No. The free self-service tier closed in July 2026. Enterprise access and Amadeus's partnership programmes remain available, both through a commercial conversation, and Amadeus doesn't publish pricing.

Can I integrate Sabre without IATA accreditation? 

You can test against Sabre's free sandbox without it. Issuing real tickets requires accreditation, a consolidator or a host agency, plus a paid production agreement.

How long does GDS integration take? 

Roughly 24 to 60 weeks for direct access, covering commercial approval, development and certification. Aggregators typically take 4 to 16 weeks. These are integrator estimates.

Do I need more than one GDS?

Rarely at launch. Most platforms start with one and add supplementary sources for low-cost carriers or hotels.

What's the cheapest way to start selling flights? 

An aggregator with managed accreditation. No upfront cost, no IATA requirement, and a per-booking fee of around $7 on a $400 ticket at Duffel's published rates.