Most people who reach this page are looking up one address, /i/api/search/dynamic/flights/poll, which sits behind the Kayak flight results. This page explains what that endpoint is and what the Kayak rules say about it. It then shows the official KAYAK API and measures what Kayak reads from an IP address. The short answer is that, for Kayak data, a proxy is rarely the right tool.
What the Kayak flights poll endpoint is
As its name suggests, it is the address that the Kayak results page polls while flight results come in. It lives under /i/api/, the internal part of the site, and the Kayak developer portal does not list it, because it is not a public API.
The rules around it are plain. The robots file of kayak.com lists "Disallow: /i/api/search/dynamic" for every crawler, in the group for all user agents and again for Googlebot. The Kayak terms, last updated on 17 August 2026, forbid users to "Scrape, harvest, deep-link to, train AI on, or otherwise use automated tools on our Site without our written permission." They also forbid users to "Bypass any security feature, access restriction, or robot exclusion header."
The robots file, generated on 25 September 2026, draws a finer line than most. It allows crawlers to read the hotel results endpoint, /i/api/search/v1/hotels/poll, and closes the flight one. Its group for ChatGPT-User closes all of /i/ while it opens the public flight, hotel and car pages.
We sent one plain request to the flight endpoint from our own server on 27 September 2026. It answered HTTP 405, with an error page titled "Unexpected Turbulence", so the endpoint does not answer a simple read at all.
The official route: the KAYAK API
Kayak does sell its data, through the KAYAK Affiliate Network. Its developer portal lists these APIs:
| API | what it returns |
|---|---|
| Flights Search | flight results |
| Flights Price Insights | the cheapest prices seen last by Kayak travelers |
| Hotels Search | hotel results |
| Cars Search | rental car results |
| Static Data Feeds | static data for hotels and places |
| Autocomplete | place suggestions |
| CompareTo Ads and Inline Ads | Kayak advertisements |
Access runs through an application. You describe your business and use case, and you can ask for free sandbox keys that return test data. Production keys follow once Kayak approves the integration. Each key is sent by email, and it goes in the apiKey parameter of every request.
The documentation also sets three rules for every call:
- One identifier per real user. Search APIs need a userTrackId, "a unique identifier per end user". Getting it wrong "can lead to rate limiting and blocking of requests."
- The real client. The User-Agent header "must always identify your actual requesting client and platform."
- The real IP. The x-original-client-ip header should carry the IP address of the user, which Kayak says "is used to identify and take measures against bots."
So the API needs no proxy. You call it from your own server with your key, and pass the IP address and User-Agent of each user in the headers. The address a call leaves from does not change those headers. When a partner cannot get the IP address of a user, the documentation says this requirement can be lifted.
Before you apply, have three things ready: a short description of your business, the use case you plan to build, and the site or app that will show the results. The partner form asks for exactly these, plus whether you want sandbox access. The sandbox has lower rate limits than production, and Kayak says the limits "may change at any time".
What Kayak reads from an IP address
We opened the front pages of four Kayak country sites from our server in St. Louis on 27 September 2026, as plain requests with a browser user agent. The last column shows the departure airport each page filled in.
| site | currency, language | airport |
|---|---|---|
| kayak.com | USD, en-us | STL |
| kayak.co.uk | GBP, en-gb | STL |
| kayak.de | EUR, de-de | STL |
| kayak.co.jp | JPY, ja-jp | not checked |
Two things follow. First, the market follows the domain, not the IP address. A US address got British pounds on kayak.co.uk and euros on kayak.de, with no redirect to kayak.com. So you do not need a proxy to see the currency and language of another market; open its Kayak site. We did not compare fares.
Second, the IP address sets the starting point. Every page we opened carried an ipLocation object naming St. Louis, the city of our server, and it filled in St. Louis as the departure airport. Since the page reads that city from the address, an address in another city should name that city instead; we did not test one.
When a proxy helps, and when it does not
It helps with one narrow job: seeing the front page the way a traveler in another city sees it, with their airport filled in. Our Premium residential traffic targets cities, at $1.49 per GB from the first gigabyte.
It does not help with the rest:
- Other markets. Open the Kayak site of that country; the domain sets the currency and language.
- Data at scale. The terms forbid automated tools without written permission, and the robots rules close the endpoint. Apply for the API instead.
- The API itself. It needs no proxy: call it from your server with your key, and pass the IP address and User-Agent of each user in the headers.
If you track fares on sites that allow it, our travel fare aggregation page covers that work.
For other work, where a proxy is allowed, start with what a residential proxy is and when to choose rotating or static residential proxies. Our free proxy list and proxy checker cost nothing. On sites that allow automation, our guides for Playwright and Puppeteer show how a scripted browser takes a proxy. We sell residential, ISP and mobile proxies. Residential traffic without city targeting starts at $0.44/GB, and every plan is on our pricing page.
The limits of this page
We made plain requests without JavaScript or cookies. So we did not see a search run, and we did not test how Kayak treats repeated or automated traffic. A script also shows a different TLS fingerprint from a browser, so Kayak may answer a browser differently. We checked currency and language, not individual fares, so this page does not say whether a fare changes with the IP address of the visitor. The API reference beyond the Getting Started text was not read. The terms and the robots rules can change, and this page is due for a check by 27 December 2026.
Sources
Every source below was read on 27 September 2026.
- KAYAK, Terms of Use, last updated 17 August 2026.
- KAYAK, robots.txt.
- KAYAK Affiliate Network, the network page and the API page.
- KAYAK for Developers, the developer portal and its Getting Started text.
- Our requests from our own server in St. Louis to kayak.com, kayak.co.uk, kayak.de, kayak.co.jp and the poll endpoint, 27 September 2026.
More pages in this series: proxies for Google Flights, proxies for Skyscanner, proxies for Booking.com, proxies for Agoda and proxies for Priceline.
