Kijiji Real Estate Scraper (Canada)

Scrapes real estate listings from Kijiji — Canada's largest classifieds site.

Public data only. No login, no cookies, no browser. No bot challenge on any of 8 TLS profiles tested across search and detail pages.

How input works: paste a search URL, don't type a city name

Kijiji's location ids are opaque integers (Toronto = 1700273, Vancouver = 80003, ...) with no public autocomplete to resolve a name into one. Rather than guess, this actor takes full search URLs — build your query in Kijiji's own UI (any location, category or filter combination) and paste the resulting URL:

https://www.kijiji.ca/b-real-estate/toronto/c34l1700273
https://www.kijiji.ca/b-for-rent/vancouver/c30349001l80003

Ignore Kijiji's own ?page=N in a copied URL. It does nothing — page 2 of a query returns the identical page-1 payload. Pagination happens through a different mechanism this actor handles for you (a page-N path segment), so any query string on a pasted URL is stripped rather than trusted.

What you get

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

PROPERTY — one row per listing

Upstream's listing object, passed through verbatim: title, description, price, location (address, nearest intersection, GPS coordinates), image URLs, ad source, activation date. With Fetch full property details on, each row also gets a detail object: canonical attribute names and values (bedrooms, bathrooms, unit type, amenities...), MLS data where the listing is MLS-sourced, virtual tour and video links, and the request-viewing URL.

SEARCH_SUMMARY — one row per search URL

Kijiji's own match total, pages fetched, requests spent.

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	One Kijiji search URL per entry.
Fetch full property details	One extra request per listing. Off by default.
Max listings per search URL	0 for everything Kijiji will serve.
Max pages per search URL	0 for no limit — the actor stops correctly at the true end regardless (see below).
Max concurrent requests / Minimum seconds between requests	Throughput controls.

Known limits

• URLs, not location names. See above — this is a deliberate choice, not a missing feature; there is nothing reliable to resolve a name against.
• A location id Kijiji cannot resolve is an honest empty result, not an error and not a silent fallback to nationwide results — totalMatches: 0, zero rows, zero errors.
• Real estate only. Kijiji is also a huge general marketplace, jobs board and vehicle listings site on a similar contract — out of scope here.
• Pagination clamps past the true last page rather than going empty — the site itself re-serves page 1 if you ask for a page number past the end. This actor detects that (stopping when a page carries no listing ids it hasn't already seen) so it is invisible to you, but it means "keep paging until the page is empty" would be the wrong approach for anyone extending this actor.