Zigbang Property Scraper

Scrape Korea's largest home-finding marketplace: oneroom (원룸), villa (빌라) and officetel (오피스텔) listings across Seoul and the rest of the country, with deposits, monthly rents, floor info, addresses, subway access and public agent contacts.

Why this actor

• Search by location: resolve a subway station, neighbourhood or district (강남역, 홍대입구, 마포구...) and read every listing around it, with server-side deposit, rent, maintenance-fee, size and deal-type filters.
• URL mode: paste listing URLs or bare listing ids and read exactly those.
• Detail enrichment: an optional per-listing pass adds the full description, options (appliances and furniture), elevator, bathroom count, direction, parking, move-in date, nearby subway lines and the agent's public contact block.
• Resume and incremental monitoring: continue an interrupted pull, or schedule the same search daily and get only NEW / UPDATED / EXPIRED rows.
• Structured output: flat records with a stable field contract (see the Output tab views).

Quick start

Search mode (default). The keyword resolves through the source's own location search; when it names a subway station the actor reads everything within the radius, otherwise the area around the resolved location:

{
  "mode": "search",
  "query": "강남역",
  "serviceType": "oneroom",
  "subwayRadiusKm": 2,
  "maxItems": 50
}

Narrow with deal types and price bounds (bounds are in 만원 units, i.e. 10,000 KRW):

{
  "mode": "search",
  "query": "홍대입구",
  "serviceType": "oneroom",
  "salesTypes": ["jeonse", "wolse"],
  "depositMax": 2000,
  "rentMax": 100,
  "maxItems": 50
}

URL mode: paste listing URLs or bare ids from any listing type:

{
  "mode": "url",
  "urls": ["https://www.zigbang.com/home/oneroom/items/12345678", "12345679"]
}

URLs in this README are fictional examples shaped like the real ones.

Deal types

value	Korean	meaning
jeonse	전세	lump-sum key-money lease
wolse	월세	deposit plus monthly rent
maede	매매	purchase
dongi	단기	short term

Oneroom listings on the source have no sale (매매) type; sale listings live in the villa and officetel verticals. Apartment-complex (아파트) resale listings are not part of this actor's scope today.

Filters

The filter section (deal types, deposit, rent, maintenance fee, size) applies in search mode only; a pasted listing is read as it is. All filter bounds are optional and sent to the source server-side, so a filtered run reads exactly the matching slice.

Output

Every record carries the listing card: id, url, title, deal type, deposit, rent, maintenance fee, sizes, floor, address parts, coordinates, registration date, tags and thumbnail. With fetchDetails on, each record also carries the description, options, elevator, bathroom count, direction, parking, move-in date, view count, nearby subway lines and the agent block (office name, public office phone, office address and registry number). Agent contact data is public business contact information the source itself publishes on every listing page.

Incremental runs add changeType (NEW / UPDATED / UNCHANGED / REAPPEARED / EXPIRED), changedFields, firstSeenAt and lastSeenAt.

Resume and recurring updates

• Resume: paste a previous runId or dataset id into resumeFromRunId; records already collected there are skipped.
• Incremental: set incrementalMode and schedule the same search; the actor remembers its own baseline (keyed on the search scope) and classifies every listing. Suppressed unchanged rows are not returned and not billed; emitUnchanged and emitExpired return (and bill) those extra rows.

Send results into your apps (MCP connectors)

Set mcpConnectors to one or more connector ids and the run's records are exported through the Apify MCP proxy after the dataset is written (Notion pages with notionParentPageUrl; at most maxNotifyListings records). Export is best-effort: a failing connector never affects the scraped data.

Notes and limits

• The source's robots.txt asks crawlers to stay out except named search engines; this actor reads only the public listing data the site's own web app serves to every visitor. Respect the data and any applicable terms in what you do with it.
• One search reads one location scope (a subway radius or one geohash cell, roughly 39 x 20 km). Split city-wide pulls into district-level searches.
• Listing cards come in batches; a run under a heavy network failure that cannot read any batch fails loudly instead of returning an empty success.