Argenprop Real Estate Scraper (Argentina)

Scrapes property listings from Argenprop — one of Argentina's largest real estate portals (188,801 results in "Departamentos en venta" alone).

Public data only. No login, no cookies, no browser.

The one thing you need to know before using this

You paste a search URL — you don't build one. Filters (property type, operation, neighbourhood, price, bedrooms) are applied through Argenprop's own search UI and end up encoded in the URL path (/departamentos/venta/palermo), not a query string. There's no public directory mapping place/type names to Argenprop's internal path segments, so this actor never guesses one — browse to any search on argenprop.com and paste the resulting URL in. Any query string on the pasted URL is dropped: Argenprop's robots.txt only permits a bare ?pagina-N query string (see Known limits), so a filter carried as a query param wouldn't survive pagination anyway.

What you get

Two record types share one dataset, told apart by recordType.

PROPERTY — one row per listing

Search cards are already rich — parsed straight from Argenprop's own inline data-* attributes, no JSON blob or JSON-LD needed: price and priceNormalized (raw integers, Argenprop's own currency-normalized figure — no locale-format "2.400.000" string parsing involved), priceCurrency, address, locationLabel, bedrooms, expensesText (monthly HOA-style fee), a features list (variable per listing — surface area, age, bathrooms, garage, whatever that card actually shows, captured as {icon, text} rather than assumed fixed), photo count and URLs, agency name and logo, and Argenprop's own internal taxonomy ids (propertyTypeId, operationTypeId, provinceId, partidoId, localidadId, barrioId, subBarrioId) — no public id→name lookup exists for these, so they're passed through raw rather than guessed at.

SEARCH_SUMMARY — one row per search URL

Pages fetched, properties returned, and (best-effort) Argenprop's own total match count from the page's <h1> when it renders one.

ERROR — one row per input that could not be processed

Every input maps to at least one row, so nothing disappears silently.

Input

Field	What it does
Search URLs	full argenprop.com search-result URLs, copied from the site — one crawl each
Max properties per search URL	0 = unlimited, still capped at 200 (10 pages × 20/page — see Known limits)
Max concurrent requests	absorbs latency variance across search URLs — not the throttle
Min interval between requests	the actual speed control — 1.0s default

Example

{
  "searchUrls": [
    "https://www.argenprop.com/departamentos/venta",
    "https://www.argenprop.com/casas/alquiler"
  ],
  "maxItems": 200
}

Notes on reliability

• AWS WAF rate/volume gate, not a per-request block. The first several dozen requests from a given exit IP go through clean (200, real content); past that budget, every response becomes HTTP 202 with an window.gokuProps challenge shell. Confirmed IP-based, not TLS-fingerprint-based. Residential proxy IP rotation on retry is the real mitigation — this is why proxy defaults to Residential/Argentina rather than being merely a defensive default, and why the default request interval is more conservative (1.0s) than most of this portfolio.
• List-only — no detail-page pass. The WAF gate tripped during recon before a detail page's shape could be verified, so this actor does not guess at one. The search card already carries price, address, bedrooms, expenses, photos and agency — enough for a browse without an extra request per listing.
• Speed is controlled by request rate, not concurrency, same design as the rest of this portfolio — maxConcurrency absorbs latency variance, minRequestInterval is the actual throttle.

Known limits

• 10-page policy ceiling (200 listings) per search URL — from robots.txt, not a technical limit. Argenprop's own Allow/Disallow rules permit only the bare query string ?pagina-1 through ?pagina-10, end-anchored; anything else matching pagina- is disallowed. This actor respects that even though the WAF's own technical ceiling sits well higher — narrow your search (by neighbourhood, price, etc., all via the URL path) rather than requesting more pages of a broad one.
• No location/type builder. There is no authoritative public list of Argenprop's path segments for property type, operation or neighbourhood (the on-site location autocomplete posts to a separate API domain, api.sosiva451.com, not used here) — see "the one thing you need to know" above.