Kayak Flights Scraper

Scrape Kayak flight result pages into clean itinerary records with prices, booking providers, airline-direct flags, baggage signals, fare options, legs, segments, aircraft, and cabin details.

The actor drives Kayak's own flight results page in Playwright and captures the signed dynamic flight poll responses that the page emits. It can also paginate through the same endpoint after the first page is captured.

What You Get

• Flight itinerary price, currency, provider, booking link, rank, and Kayak share URL
• Airline names and codes across all legs
• Per-leg and per-segment times, airports, duration, aircraft, and flight numbers
• Booking options with direct-airline flags, free-cancellation flags, carry-on and checked-bag information
• Fare buckets where Kayak returns them, such as basic economy, main cabin, or business fare variants
• RUN_SUMMARY and OUTPUT records in the key-value store for automation-friendly status checks

Input

Use either structured route fields or full Kayak URLs.

Structured example:

{
  "origin": "LAX",
  "destination": "NYC",
  "departureDate": "2026-09-15",
  "returnDate": "2026-09-22",
  "adults": 2,
  "cabinClass": "business",
  "sortMode": "price_a",
  "maxResults": 25
}

URL example:

{
  "startUrls": [
    "https://www.kayak.com/flights/LAX-NYC/2026-09-15/2026-09-22/business/2adults?sort=price_a"
  ],
  "maxResults": 50
}

Structured searches accept airport or metro codes such as LAX, NYC, LON, and SFO. For custom Kayak searches, paste a full Kayak /flights/... result URL into startUrls.

Pricing

This actor uses pay per event:

• apify-actor-start: $0.00005 per GB of run memory, minimum one event
• flight-found: $0.006 per validated flight itinerary stored in the dataset

Filtered-out, invalid, blocked, or empty results are not charged as flight-found events.

Notes

Flight prices and availability change frequently. The dataset reflects what Kayak returned during the actor run. Residential proxy is recommended for cloud runs; direct local runs may work for probing, but datacenter IPs can be blocked.