Choosing a flight booking API provider: look behind the API

The shortlist usually looks good. Three flight booking API providers, each with clean REST documentation, a sandbox key in the first week, and a search endpoint that returns fares for the routes you tried. The team picks the one with the nicest SDK, builds search, pricing and booking in a sprint or two, and schedules a demo. Then, close to go-live, somebody asks who actually issues the ticket. Sometimes the answer is that you do, which needs an accreditation the company does not hold, or a deposit nobody budgeted for. Sometimes the ticket issues fine, and the first refund request turns out to be an email to a support desk in another time zone.
None of that was in the docs, because it is not a documentation question. Picking a flight booking API provider is mostly picking what stands behind the API: whose ticketing authority issues the e-ticket, whose money moves and when, and who answers when a ticket has to be voided, refunded or reissued. Search, price and book look roughly the same everywhere. That is the part teams compare, and it is the part that is the same.
Four kinds of provider sell the same endpoints
From the outside, every flight API has a search call, a price or offer call, a booking or order call, and some way to get to a ticket. Behind those calls sit quite different commercial chains. Roughly four kinds of provider turn up on a shortlist, and the categories blur at the edges, because plenty of companies are two of them at once.
| Provider kind | Who issues the ticket | What you need before you can sell | How money moves | After the ticket |
|---|---|---|---|---|
| A GDS, directly | Usually you, on your own ticketing authority | Typically an accreditation and a ticketing arrangement of your own, plus the commercial agreement with the GDS | You settle with the airlines through the settlement plan your accreditation gives you (BSP in most markets) | Rich servicing commands exist, but your team is the one running them |
| An airline's NDC channel, directly | You, or the airline against your agreement with it | An agreement with each airline, one airline at a time | On each airline's terms, which may or may not run through the same settlement plan | Changes and refunds where that airline has built them; it varies carrier by carrier |
| An aggregator over GDS and NDC content | Depends on the aggregator: you on your authority, or the aggregator on its own | Ask. Some need your accreditation, some do not | Ask. Deposit, credit or card per booking | One normalised interface, but each operation is only as complete as the source behind it |
| A consolidator or B2B platform API | The consolidator, on its ticketing authority | An account with the consolidator; no accreditation of your own | A deposit or credit line with the consolidator, settled on its terms | Whatever the platform exposes as calls; some operations may still be requests its desk handles |
You can argue with any cell, and you should, with the provider across the table. The row that matters is not the one with the longest feature list. It is the one whose first three columns describe the company you already are. A startup without accreditation that signs with a provider which assumes one has built against the wrong row, however good the SDK.
Ask a flight booking API provider these before the contract
The answers are rarely on the docs site. They come out of a commercial conversation, and that is not a warning sign: a provider that issues tickets on its own authority has to know who you are before it lets you sell on it. Our piece on what a free flight API actually gets you makes the narrower point that bookable inventory always has somebody carrying the ticket behind it. These are the questions that find out who.
- Whose ticketing authority issues the ticket? If yours, what accreditation and deposit does that take in your country, and how long does it take to get. If theirs, whose name the airline sees on the booking, and whose name your customer sees.
- How is money held and settled? A pre-funded deposit, a credit line, a card charge per booking. And the follow-up nobody asks: what happens to a booking request at four in the afternoon when the deposit has run out.
- Which after-ticket operations are API calls? Void, refund, exchange, handling an airline schedule change, a name correction, adding a bag or a seat. For each one, write down whether it is an endpoint, a request a human handles, or not offered at all. Do it operation by operation, not as a yes to "servicing".
- What does coverage look like from your point of sale? Ask for results on your own top routes, searched from your own market, not on the demo route. Content and fares differ by where the sale is made.
- How far does the sandbox go? Search and price are easy to fake. A sandbox that goes as far as hold, ticket, void and refund in test is testing the half that breaks in production.
- Who picks up when a ticket is in a strange state? A named desk with hours that overlap yours, or a ticket queue.
Once the chain fits, the engineering questions start to matter: prices that move between search and booking, a PNR versus an order, timeouts that leave you unsure whether a booking exists. Those are covered in what the flight API docs leave out, and our travel API comparison gives a way to weigh the technical axes against each other. Neither is worth doing until the row in the table is settled.
Where this platform sits
We sell flight access through an API, so here is our row, stated plainly. The platform exposes flight and hotel APIs. Access starts as a conversation rather than a self-serve key, because the first questions are the commercial ones above, and they are answered per company. Services are activated per tenant. Consolidator accounts exist on the platform, and sub-agent credit limits and settlement invoicing are first-class concepts in it rather than spreadsheets.
We do not publish carrier coverage or prices on this page. Both depend on your market and your point of sale, and a number written here would be wrong for most readers. If you want to put the list above to us, start with the travel APIs page.
What choosing on the docs costs
Two outcomes, both expensive, in different currencies.
The first is an integration that works and cannot sell. Search, price and book pass in the sandbox; in production the booking cannot be ticketed in your market, because the ticketing arrangement the provider assumed does not exist on your side. The code is finished and waits on paperwork that should have been the first question, and the paperwork has its own timetable.
The second is quieter and lasts longer. Tickets issue, the product launches, customers arrive. Then every void, every refund and every schedule change becomes an email thread with the provider's desk, so your support headcount grows in step with sales and your margin per ticket shrinks with it. The API looked complete because the happy path was complete. Nobody demos a refund.
Neither failure shows up in a proof of concept. Both show up in the first month of real customers. So read the docs last. Draw the chain first: who issues, who pays whom, who services what. Pick the provider whose chain matches the company you are, and only then argue about SDKs.