Africa's largest classifieds network as JSON, in five countries

Five: Nigeria, Ghana, Kenya, Tanzania and Uganda, passed as country with the values ng, gh, ke, tz and ug. Nigeria is the default. Each one is a separate storefront with its own catalogue, its own region tree and its own currency, so a category slug or a region slug from one country is not guaranteed to exist in another. The free categories and regions actions list exactly what that storefront publishes.

Here, yes, and it was measured rather than assumed. Jiji itself accepts a filter name it does not know and hands back the unfiltered list, so a query that looks like it worked can silently be the whole category. Every filter you send is checked against the schema Jiji publishes for that category before any search is issued, and a name or a value the schema does not carry comes back as an error naming the ones that do. 23 filter cases were run against a same-run unfiltered control and all 23 narrowed: make=Toyota took 59,504 car ads to 24,730, make plus model to 6,349, a 2018-2021 year range to 7,474, a mileage ceiling to 2,603, a price band to 850, a region to 41,601 and a dog breed from 3,713 to 4.

Ask the free filters action with a category. It returns every filter that category publishes, with its key, its label, whether it is single-choice, multi-choice or a range, its unit, and the complete list of allowed values or the range bounds. Some filters only exist once a parent choice is made: the model list appears after a make is chosen, so pass the parent back into filters to see the children it unlocks. In search, filters are name=value pairs separated by a semicolon, ranges take from-to, from- or -to, and several values for one filter are comma-separated: make=Toyota;model=Camry or condition=Foreign Used;year_of_manufacture=2018-2021.

Because Jiji publishes none for them. Job ads are the clearest case: 0 of 23 general job rows and 0 of 24 IT job rows carried a price, and 18 of 23 cleaning-service rows carried none either. Those ads come back with price null and price_not_published true, plus Jiji's own price_label kept verbatim, so a missing price is never reported as 0 and never reads as free. In the other 15 categories measured, every row in the sample carried a price.

The listing action returns the phone number or numbers the ad itself publishes, in seller.phone and seller.phones, and it did so on 22 of 22 detail calls across 13 verticals. It is on by default and include_phone set to false turns it off, which also saves one upstream call. Jiji publishes no seller e-mail address to a logged-out reader, so no e-mail field is returned: 0 hits in 30 ads scanned end to end and 0 in 38 raw payloads. The rest of the seller block is the public profile Jiji shows anyway: display name, account id, rating, profile or shop URL, avatar, advert and feedback counts, how long they have been on Jiji, when they were last seen, and their trust labels.

Jiji's page is fixed at 24 rows and cannot be widened, so limit above 24 is served as consecutive pages, up to 96 rows in one call, with duplicates dropped. The search window itself is Jiji's: a 58,000-ad category reported 223 pages, which is roughly 5,350 ads reachable per query. Asking past total_pages is an error rather than an empty page. To go wider than the window, partition the query by region, price band or filter, which is what the free regions and filters actions exist to give you.

Whatever Jiji does not publish, reported as absence rather than guessed. Four categories publish no row-level attributes at all, and not one row in the samples carried any: cleaning-services 0 of 23, babies-and-kids-accessories, dogs-and-puppies and livestock-and-poultry 0 of 24 each, and building-materials 2 of 23, while cars, phones, property, land, laptops, furniture, appliances and jobs carry a full attribute table on every row. condition is null wherever the category defines none, which includes property, jobs and services. seller.name is attached to a search row only on some ads, from 16 of 24 on furniture down to 0 of 24 on houses for sale, but seller.user_id is complete everywhere and listing always has the name. listing_id, advert_id, title, url, image, description, location.region and seller.user_id were complete in all 18 categories measured.

It means no ad matched, and the answer says so explicitly. An empty result is re-checked before it is returned, and empty_confirmed true is only present when the re-check agreed, so a real no-match is distinguishable from a blip. A sold, expired or removed ad is a different case: Jiji serves it nothing at all, so a dead listing_id is NOT_FOUND rather than an empty success.

Jiji API is a ReefAPI endpoint group for africa's largest classifieds across nigeria, ghana, kenya, tanzania and uganda: ads in local currency, the seller and their whole stock. It returns live JSON through POST requests under /jiji/v1.

Yes. ReefAPI starts with 1,000 free credits, no card required. Jiji calls use the same shared credit balance as every other ReefAPI engine.

No login to Jiji is needed for the API response. You call ReefAPI with your x-api-key header, and the playground can run live examples before you create a production key.

The page example is captured from a live search call, and production requests fetch live data through ReefAPI rather than a static sample.

Jiji actions currently cost 1-2 credits per successful call. Failed or blocked calls are free. All APIs draw from one credit pool.

Yes. Connect ReefAPI once through MCP and your assistant can call jiji actions with the same key, credit pool and JSON envelope used by normal REST requests.