CarSensor Japan Used Cars (カーセンサー) Scraper

Pulls dealer-stock used-car listings from carsensor.net, Recruit's used-car marketplace and one of the two largest in Japan (≈500,000 cars). Filter by maker, model and prefecture; get every list-page field and, optionally, the full detail page (fees, specs, warranty and maintenance text, dealer comment, images). No login, no API key.

Who uses this: used-car exporters and parallel importers pricing against Japanese dealer stock, valuation and price-index vendors, dealers benchmarking competitors, and analysts tracking JDM supply by model and region.

Input

field	type	notes
brand	string, required	2-letter maker code from https://www.carsensor.net/usedcar/b<CODE>/: TO Toyota, HO Honda, NI Nissan, MA Mazda, SB Subaru, SU Suzuki, DA Daihatsu, MI Mitsubishi, LE Lexus, ME Mercedes-Benz, BM BMW, VW Volkswagen, AU Audi, PO Porsche …
model	string	Model code, the s<code> segment of a model URL (/usedcar/bTO/s230/ → 230, Pixis Epoch). s230 is accepted too. Codes are in the model menu of any maker page.
prefecture	string	Romanised URL slug: tokyo, kanagawa, osaka, aichi, hokkaido, fukuoka … (the site spells kouchi, ooita). Regions work: hokkaido, tohoku, kanto, koshinetsu, tokai, kansai, chugoku, shikoku, kyushu.
maxItems	integer	Stop after this many cars (default 100, max 30,000).
fetchDetails	boolean	Open each car's detail page (default true). Off = list data only, ~30 cars per request.
maxConcurrency	integer	Parallel requests, 1–20 (default 5).
proxyConfiguration	object	Optional; the site answers plain requests.

Example: { "brand": "TO", "model": "230", "prefecture": "kanto", "maxItems": 200 }

Output (one record per car)

From the list page: id (AU…), url, brand, title, subText (dealer's one-line pitch), bodyType, color (short), totalPriceYen (支払総額, all-in), basePriceYen (車両本体価格), planPrices (A/B プラン totals), monthlyLoanYen, year + yearJp (2022 / R04), mileageKm, inspection (車検), repairHistory / hasRepairHistory (修復歴), warranty, maintenance, displacementCc, transmission, labels (360°画像付, 車両品質評価書付, 販売店保証 …), prefecture, city, shopName, shopUrl, shopReviewScore, imageUrl.

Added by the detail page: name, expensesYen (諸費用), warrantyDetail, maintenanceDetail, oneOwner, nonSmoking, serviceRecords, unregisteredNew, driveType, colorName, steering, seats, fuel, doors, shopComment, images (main photo from the page's structured data; the gallery is JS-loaded), and specs — every row of the 基本情報 / 車両詳細 / 型式スペック tables as Japanese key → text (wheelbase, dimensions, weight, fuel economy, chassis-number tail, …).

Prices are integers in yen (the site shows 万円: 59.9万円 → 599000). null means the site did not show the value (e.g. 応談 = negotiable).

Pricing

Pay per event: one car event per record pushed. Set a maximum total charge on the run to cap spend; the actor stops cleanly at the budget or at maxItems, whichever comes first, and never charges for a car twice — even across migrations and restarts.

Notes and limits

• The site lists 32 cassettes per page, of which 1–2 are lease ads (/usedcar/lease/); those are skipped, so pages yield ~30 cars.
• Prefecture-filtered lists show a site-side count that tops out at 2,000; a maker-wide list (e.g. Toyota, ~118,000 cars) paginates much deeper. Narrow by model or prefecture for exhaustive pulls.
• Past the last page the site serves page 1 again; the actor detects this from the pager and stops.
• An unknown maker code fails the run immediately; an unknown model or prefecture path yields the site's empty list.
• Sold or removed cars answer 404 on the detail page and are skipped without charge.
• Keep concurrency modest; the actor sends plain GET requests with no proxy by default.

Terms

carsensor.net is a public site; the actor reads only pages that are visible without an account and respects robots.txt (no Disallow). The site's 利用規約 (car-t-1008) forbids obstructing the service's operation; the default concurrency of 5 keeps load light. Data is collected for the user's own analysis; dealer names and phone numbers shown on the site are business, not personal, information. Use of the data remains the user's responsibility.