Every second-hand copy in the Low Countries, as one JSON API

On the book page, yes, and exactly as the seller printed it: 11 of the 12 books we audited carried one. On a search row, no — and we would rather say so than invent it. Boekwinkeltjes' result table has seven columns (image, author, title, publisher, particulars, price, shop) and no ISBN cell, so isbn is null on all 250 rows we counted. The useful direction still works: pass an ISBN as query and the site matches on it — 9789022537510 returned 7 copies — then call book on the ids you care about. ISBN-10 and ISBN-13 are returned in separate fields by the length the source printed, and neither is ever converted into the other, because a computed check digit would be our number and not the seller's.

A copy. That is the whole point of this source. Every row is one physical copy, in one shop, at one price. In a single live result page, book 219742668 and book 244766468 both carry ISBN 9789027401700 — the same edition — at EUR 27.50 and EUR 5.30 from two different sellers. So a title query returns as many rows as there are copies on shelves across the market, and total_estimate counts copies. If you want the cheapest copy of an edition, search its ISBN and sort by price.

In one free-text field the site calls Bijzonderheden, which we return verbatim as particulars — for example '2024 256pp Gebonden' or 'paperback, 1975, eerste druk, 201p., 20,5 x 13,5 cm'. Boekwinkeltjes has no structured year, edition, page-count or binding field, so we do not pretend to have one: condition and binding are filled only on rows where the seller literally wrote 'Conditie:' or 'Bindwijze:', which was 34 and 26 of 250 rows. Everything else stays in the text, where you can read it, rather than being guessed out of prose and shipped as if it were a field.

It is checked against the source twice. The book page prints its price in its own detail table and again in its own structured data, and we compare both against the price on the search row. On 12 books that was 12 out of 12 agreement with zero mismatches, on two separate runs. We also return the site's own printed string next to the parsed number — '€ 27,50 (Excl. verzendkosten)' beside 27.5 — so you can always see what we read, and if the two witnesses ever disagree the response says so instead of silently picking one. Dutch notation is handled explicitly, which matters: a comma is the decimal separator here, so a careless reader turns 627,30 into 627300.

Yes, and in full on the book page: shop name, the town it sits in, whether Boekwinkeltjes classes it as a business or a private seller, its slug, its own website where it has one, its logo, how many working days it says it needs to get back to you, and its delivery terms in its own words. On the 12 books we audited, name and business-or-private were present 12 of 12, town 11 of 12, terms 8 of 12. On search rows the shop name is there on 165 of the 208 marketplace rows. There are no seller ratings, review counts or sales totals on this marketplace — not hidden, simply not published — so those fields do not exist rather than coming back empty.

Boekwinkeltjes mixes in stock from a commercial partner, and those rows sit at the top of result pages. We measured 42 of 250 rows, about one in six. Their shop cell is a button image instead of a shop name, so seller_name comes back null on them and listing_channel says partner_boekenbalie, and every search reports how many of your rows were partner rows. We keep them because they are real books with real ids that resolve to a full record — where the seller IS published — and dropping a sixth of the market without telling you would be worse than labelling it.

Ask for total_estimate and compare. We did, in a single run against an unfiltered 2,429 copies for q=tolkien: second-hand only 1,983, new only 446 (and 1,983 + 446 is exactly 2,429, so those are the two complete halves), books in English 801, in German 61, sellers in Belgium 470, sellers in the Netherlands 2,393, at least EUR 50 gives 183, at most EUR 5 gives 376, only with a photo 1,652, only listings that charge shipping 1,606, added in the last week 103, and within 10 km of postcode 1012 just 72 — against 1,673 within 150 km. All 11 exposed filters narrowed the set. This matters more than usual here, because the site itself accepts a parameter it has never heard of and returns the full unfiltered result with a 200: we checked, and an invented filter left all 2,429 in place. So every value you pass is validated on our side and a typo is rejected with the allowed list, rather than quietly doing nothing.

No, and it is the opposite, which is why we renamed it. The site's own switch reads 'show only listings with shipping costs', so our parameter is shipping_cost_listed: it keeps the copies that DO charge for postage. On q=tolkien that is 1,606 of 2,429. If you want the free ones, read shipping_eur on the rows: 'Gratis' is returned as 0, a real amount as that amount, and a copy where the seller printed nothing as null.

50 copies per page, and that was constant on every one of the 12 pages we sampled. The site's own ceiling is page 200 — page 200 returns 50 copies, page 201 returns nothing at all — so a query can reach 10,000 copies at most, whatever its total is. The API rejects a page above 200 with the reason instead of letting you discover it, and tells you in the response when you have hit the wall. To go deeper than 10,000, split the query: by language, by seller country, by price band or by postcode radius, all of which are filters here.

A query that matches nothing is an answer, not an error: you get ok with zero books and a stop reason of 'empty'. A copy that no longer exists is different — a dead book id returns NOT_FOUND, non-retryable, as does a shop slug that does not exist. Those are real answers from the source, so the API hands them back as such rather than as a vague failure you have to interpret, and never retries them.

Yes, that is the seller action. Take seller.slug from any book response, pass it in, and you get the shop's profile plus a page of its copies in the same shape as search — and, uniquely on this source, the shop's own published stock total: Kaatjes Boeken states 9,798 books. You can also pass query to search inside that one shop only, and sort by the same keys as search. It is the only count Boekwinkeltjes prints anywhere.

Boekwinkeltjes API is a ReefAPI endpoint group for the dutch and belgian second-hand and antiquarian book market as json: every copy on offer across 11,264 bookshops and private sellers, with isbn, condition text, shipping and the shop behind it. It returns live JSON through POST requests under /boekwinkeltjes/v1.

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

No login to Boekwinkeltjes 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.