Skip to content
Blog

Amadeus Self Service API alternative: a week to migrate

Tekravel Editorial
Written with the help of AI

You find out from a failing health check, or from a user. The search endpoint your flight app has called since its first prototype returns an error you have never seen, and the portal where you managed your keys is no longer there. If you are looking for an Amadeus Self Service API alternative, you are probably in that position now: a working product, a codebase shaped around one provider's requests, and a short window before your users notice that search is broken.

One fact about the vendor is worth stating, and only one. Amadeus decommissioned its Self-Service API portal on 17 July 2026, and new developers are routed to its Enterprise channel, which runs on a contract. That is all this piece will say about it. The useful question is what replaces it, and how to move in about a week without rebuilding the parts that still work.

What self-serve actually gave you

It is easy to remember the old portal as "a flight API" and go looking for another flight API. That misses what made it useful to a two-person startup or a weekend project. Three things, mostly:

  • A test mode you could reach in an afternoon. Sign up, get a key, call search. Nobody asked what your company did.
  • Pay-as-you-go pricing. You paid for what you called, so a product with a hundred users cost what a product with a hundred users should cost.
  • No sales cycle. Nobody had to approve you before you could write code.

An enterprise contract inverts all three. Access comes after a commercial conversation, pricing is negotiated, and the test environment arrives when the paperwork does. None of this is unreasonable from the seller's side; distributing airline content carries real obligations. But it changes your plan. If you were counting on swapping one API key for another, the honest answer is that most serious replacements no longer work that way, including the one this site offers. More on that below.

Three routes to an Amadeus Self Service API alternative

Strip away the vendor names and there are three kinds of replacement. They differ less in their JSON than in who you sign with and who ends up issuing the ticket.

RouteTime to first real bookingContent you can expectWho holds the ticketing plate
Direct GDS contractLongest: commercial review, certification, then the buildBroad GDS content; NDC and low-cost carriers depend on the deal you signYou, if you hold IATA accreditation; otherwise you still need someone who does
Aggregator APIMiddle: lighter onboarding, but expect a contractA mix of GDS, NDC and LCC, depending on which sources the aggregator has connectedUsually the aggregator, or a ticketing partner of theirs
A consolidator's APIOften the shortest path to an issued ticket, because ticketing is already solvedWhatever the consolidator sells: its contracted and net fares, plus what its suppliers carryThe consolidator, on its own accreditation

You can disagree with the middle column, and for your own case you should. A team that already holds accreditation and has a commercial person on staff may find the GDS route quicker than the table suggests. But most developers who built on a self-serve sandbox never held a ticketing plate, and that last column decides how much of your product still works after the move.

Here is why. Search and pricing are read operations; any of the three routes will give you those. Issuing a ticket is a financial act, done under somebody's accreditation and settled through BSP or its equivalent, and if that somebody is not you, you need an agreement with whoever it is. Teams that skip this question end up with a handsome search page and a booking button that creates a reservation nobody pays for.

Content is the other column to read slowly. GDS content and NDC content are not the same fares, and low-cost carriers often sit in neither. If your users mostly fly short-haul in a region full of LCCs, a route with excellent GDS coverage can still return thin results. We went through that distinction in NDC API vs GDS: what changed for integrators.

How access works here, and why it is not a key either

This platform exposes flight and hotel APIs. Access starts as a conversation rather than a self-serve key. We say that plainly because the worst outcome for you this week is to sign up somewhere expecting the old experience and discover on day four that a human approval sits in the way.

The conversation is mostly about the questions in the table: what you are building, who your users are, whether you need ticketing or only search and price, and which markets you sell into. Those answers decide what content you get and how bookings settle. If you want to start it, the travel APIs page is where it begins.

Wherever you go, ask the same thing in your first email: is there a self-serve test key, and what has to happen before production? A clear answer is a good sign about the rest of the integration. A vague one tells you something too.

A migration checklist, in the order things break

Most self-serve integrations had the same shape: search, price, create an order, issue, then handle whatever happens after the ticket exists. Re-test each step against the new provider as if you had never written it. The request names will differ; it is the assumptions buried in your code that bite.

  1. Search. Map your query model (origin, destination, dates, passengers, cabin) to the new request. Check how child and infant ages are passed, whether multi-city is a separate call, and what an empty result looks like. An empty list and an error are different things, and your UI probably treats them the same.
  2. Offer identity. Find out how long an offer id stays valid and whether it survives a page reload. Code that stored an offer in the session and priced it ten minutes later will fail in new ways.
  3. Price. Re-pricing before booking is not optional with most providers. Handle the case where the price moved, and decide what your user sees when it does.
  4. Order. Passenger names, document fields, contact details. Validate formats on your side before you send them; an order rejected after the customer has paid is the most expensive bug on this list.
  5. Ticket. Confirm whether the order call also issues, or whether issuance is a separate step with its own time limit. Know whose plate the ticket goes out on before you take anyone's money.
  6. Post-ticketing. Cancellation, void, refund, schedule change. These are the calls a prototype never needed and a live product needs in its first month. Ask whether they exist in the API or go through a support desk.

For a fuller test plan, including what to fake and what never to fake, see our sandbox testing plan for travel APIs. If you are coming at it from the integration side, flight booking API integration walks the whole flow.

What a rushed migration costs

The tempting move is to wrap the new provider behind your old client interface and ship as soon as search returns results. Search will look fine. What fails is everything the demo did not show.

A price that changed between search and order, shown to the user as a raw error, loses the booking. An order created without issuance, because the new provider splits the two steps and your code does not, leaves a passenger who thinks they have a ticket and does not. A refund flow that silently does nothing until someone writes to support becomes a chargeback. Each of these is survivable once. Repeated across a user base, they cost you the trust a small travel product lives on.

There is a quieter cost as well. If you choose a route because it was quickest this week and it turns out not to carry the content your users search for, you migrate twice. A day spent on the content and ticketing questions before any code is the cheapest insurance in the whole exercise.

One week, roughly

Day one goes on the questions: which route, which content, whose plate. Days two and three on search and pricing against the new provider. Day four on order and issuance, end to end, including the failure cases. Day five on post-ticketing and on reading your own logs. If the conversation with a provider takes longer than a day, start coding against their documentation in parallel; request shapes rarely change once you have seen them.

It will not feel like the afternoon it took to get your first self-serve key. It does not need to. What you are building now has to issue tickets that someone stands behind.

Tekravel Editorial

Travel technology desk

The Tekravel travel-technology desk writes for the trade: agency owners, consolidators and the developers who integrate them. Every article is checked against the platform it describes before it is published.