All notable changes to AutoScout24 Scraper are recorded here. The newest release is first.
Apify renders this as a Changelog tab on the Store page. Keeping it current is a
visible signal that the Actor is maintained, which feeds the quality score that
drives Store and MCP search ranking. Add an entry with every published build.
- New input
listingIds: AutoScout24 listing IDs or listing URLs (any number, duplicates removed). One row per listing with its current price and facts in the same fields as a search row; lookups are not cut by maxItems (which bounds the searches only); an unknown or removed listing is a free not_found row. Re-check the prices of listings you track.
- New input
sourceDatasetId: the ID of an earlier run's dataset (for example a search run of this Actor). Every row with a listingId becomes a listing lookup; error and not_found rows are skipped.
- New watch mode
onlyNew with monitorStoreName: a scheduled search returns only listings no earlier watch run returned. Skipped listings are never charged; a search with nothing new gives one free not_found row. New sort order age_desc (newest listings first).
- Small runs are cheaper: the request queue stays in memory, and a restarted run does not write or charge saved rows twice. Output of existing inputs, prices and event names are unchanged.
- Docs: related Actors listed; input field descriptions clarified for AI agents.
- Fixed: a listing AutoScout24 pins on several result pages of the same search is now returned once per search, not once per page (you were charged for the repeats).
- Results whose text mentions "captcha" or "access denied" (a server description, a game or shop product name, a headline) are no longer mistaken for an anti-bot block, which could fail a good run.
First public release.
- Used-car listings from the AutoScout24 search pages of every country site (autoscout24.com, .de, .at, .it, .fr, .es, .nl, .be, .lu, .pl), 20 listings per request, with no login and no third-party API key.
- Search by pasted URL or by make, model, country, price, first registration, mileage, fuel, body type, seller type and sort; follows the site's pages up to its 200-page limit.
- An unknown make or model returns one unbilled
not_found row instead of unrelated cars. Private sellers' names, phones, e-mails and street addresses are never extracted.
- Every row carries
status, so a missing record (not_found) and a failure (error) are told apart from real data. Only ok rows are charged.
- Records are validated against a schema before they are written, so an upstream change produces loud warnings instead of silently billing for empty rows.
- Callable with no arguments by an AI agent, which the input schema's defaults make safe.