A book data API answers "which book is this, exactly" at scale — this one searches Open Library, the Internet Archive's open catalogue of tens of millions of editions, and returns one typed row per work: title, up to five authors, first publish year, an ISBN-13 where one exists, publishers, language codes, median page count, community rating with its count, edition count, a cover image URL and the catalogue link.
The unit is the work, and that is the deliberate part: "Dune" is one row with an edition count in the hundreds, not four hundred near-duplicate rows to dedupe yourself. Open Library's field syntax passes straight through — author:tolkien, isbn:9780261102354, subject:fantasy — and the fetch is a keyless catalogue read over the TLS tier, no browser involved, at $0.0004 per delivered book. It pairs naturally with the Wikipedia scraper API for author context and the iTunes Search API for how the same titles show up as ebooks and audiobooks.
$0.0004 per delivered book, up to 100 books per run. Nothing delivered means nothing charged.
One POST with a query. Plain text searches titles and authors the way a person would; the field prefixes turn it into a precision tool. isbn: makes it a reverse lookup — scan in hand, record out. subject: makes it a genre browser, and sort reorders the catalogue's answer four ways: relevance, new, old or rating.
Each delivered work is flattened from the catalogue's sprawling record into one row, and two normalisations do the quiet work: the ISBN field prefers a 13-digit identifier whenever the work has any, so downstream systems get the modern form; and authors and publishers are capped (five and three) so a crowd-sourced record with forty contributors cannot wreck a fixed-width import. Searches that match nothing are free.
query is required, two characters minimum, and carries the whole grammar — plain text, author:, isbn:, subject:, or combinations of them. sort defaults to relevance; rating is the one to reach for when building best-of shelves. max_results spans 1–100 with a default of 25.
rating ships with ratings_count for a reason: a 4.8 from three readers and a 4.2 from eleven thousand are different facts, and the pair lets you threshold instead of trusting an average blind. editions is an underrated demand signal — books that keep getting reprinted keep getting bought. pages is the median across editions, languages lists the codes the work exists in, and cover_url points at the catalogue's cover image when there is one.
$0.0004 per delivered book ($0.40 per 1,000). Nothing delivered means nothing charged, and the $2 monthly free credit covers roughly 5,000 books here. Volume tiers take up to 30% off.
$0.0004 per delivered book — 40 cents per 1,000. Enriching a 10,000-title backlist with ISBNs, covers and ratings is a $4 job; a bookshop's nightly new-arrivals lookup is fractions of a cent. Without a card you get $2 of monthly usage free, which is 5,000 books — many catalogue projects finish inside the allowance.
Queries that come back empty are answers, not charges, so exploratory searching costs nothing to get wrong.
Open Library's API is free, keyless and one of the most generous public catalogues on the internet — for a handful of lookups, or if you want the entire dataset (they publish full bulk dumps), go to the source. The working friction is the record shape: the raw search response carries dozens of optional fields you must know to ask for, ISBN-10s and ISBN-13s arrive mixed in one array, ratings sit behind an extra field selection, and field presence swings wildly between a modern bestseller and a crowd-catalogued 1930s pamphlet.
What this collector adds is a settled contract: one row, typed, the ISBN-13 chosen for you, ratings included, caps applied — the same discipline as every other collector on your QuantumProxies key, billed only when a book is actually delivered. The catalogue is open; the predictability is the product.
The hard problem in book data is not fetching — it is that books resist matching:
And if your pipeline also scrapes retailer sites for prices next to the metadata, the residential network underneath this collector is available for that side too.
A scanned barcode becomes an isbn: query; the row that returns carries title, authors, publisher and cover — enough to list a used book for sale or accession a library copy in one step, with no manual cataloguing.
Reading trackers, book clubs and recommendation apps need clean covers and canonical titles more than anything else. cover_url plus the work-level dedupe means one search fills a shelf view without duplicate editions.
subject: queries sorted by rating map a genre's canon; edition counts show which backlist titles still carry demand; first publish years date a trend. Raw material for acquisition memos, drawn from an open catalogue.
Metadata only — there is no full text in any row. Open Library's book lending is a separate, human-facing service, and page content is simply not what this collector reads. Rows are work-level, so you receive one representative ISBN rather than every edition's, and 100 books is the ceiling per run; wider sweeps parallelise across runs inside your plan's requests-per-minute budget.
The catalogue itself is community-maintained: coverage leans English-language, ratings are sparse outside popular titles, and the occasional record is wrong in the way open catalogues sometimes are. For corpus-scale work, Open Library publishes complete data dumps — mirroring the whole catalogue through a per-row API would be the expensive way to do a cheap thing, and we would rather tell you that here than on the invoice.
Yes — Open Library itself, which is what this collector reads. What QuantumProxies charges for is the normalised row and the fetch infrastructure: $0.0004 per delivered book, with a $2 free allowance every month that covers 5,000 books before any card is needed.
Yes. Put isbn: in front of the number — isbn:9780261102354 — and the matching work comes back with title, authors, publishers, rating and cover. Both ISBN-10 and ISBN-13 work in the query; the row you get back prefers the 13-digit form.
Every row carries a cover_url pointing at the catalogue's cover image when one exists. Works without a cover return null there instead of a placeholder, so your interface stays in charge of what a missing cover looks like.
No. The rows are catalogue metadata — titles, identifiers, ratings, covers — and never page content. Digitised lending exists on Open Library's own site for humans; programmatic full text is outside what this collector does or claims.