Ibnu Adzim

All HTTP-only, no browser, no login, no captcha — public data, read the way the site's own frontend reads it.

---

Misspell a city on Trulia and it does not error. It answers HTTP 200 with all 126,601 California listings — real homes, wrong place, nothing in the rows to warn you.

Every Actor here checks the site's own echo of what it searched:

• locationApplied: false — site ignored your query, run stops and says what it actually searched.
• resultCapReached — site holds more than it hands out; you get the real number.
• paginationClamped — page loop caught before it duplicates your dataset.

A typo costs you one row, not a dataset you can't trust.

---

• recordType, _source, _scrapedAt on every row, no exceptions
• Search + detail in one run — one toggle, not two Actors glued together
• Raw passthrough — upstream fields verbatim, never pre-flattened
• SEARCH_SUMMARY row per query, ERROR row per bad input — the run keeps going
• Declared output schema, so dataset preview and CSV match

---

Real estate (Zoopla, Otodom, Immoweb, Hemnet, Boligsiden, SUUMO, Trulia, PropertyGuru, Rumah123, ZonaProp...), Jobs (Reed, Dice, hh.ru, JustJoin.it, Bayt, JobStreet...), Travel reviews (Booking, Agoda, Trip.com, Viator...), News (Bloomberg, Euronews, Techmeme...), plus e-commerce, finance and business directories.

The pattern: the non-US long tail. Everyone builds the tenth Zillow scraper — almost nobody builds the first Boligsiden one.

---

No logins, no captcha solving, no human-session cookies, no invented fields. If a site needs any of that, it's archived, not shipped with an asterisk.

Need a site that isn't listed? Open an issue with the site and fields you need — the more obscure the market, the more likely I'll take it.