A discovery library for the unseen — built in the open, with its sources showing.

Libra Esoterica is a discovery library: a place to find out that a book exists, what it is about, who wrote it, which traditions it touches, and where you can read or buy it. It is not a bookstore, not a digital archive of its own, and not an authority on the truth of anything the books claim.

The catalog covers roughly 24,019 books across 33 traditions — alchemy, grimoires, witchcraft, astrology, Kabbalah, Spiritualism, psychical research, UFO literature, cryptozoology, mysticism, and more. Around them sits a small community layer: ratings, reviews, tags, reading lists, and shelves for readers who want to keep track of what they have found.

We treat the whole field as a subject of study. Category membership means relevance, not endorsement, and the site does not editorialize about beliefs. A skeptical survey of mediumship and a Spiritualist manual can share a shelf here because a reader of one may well want the other.

Records are harvested from 36 sources — national libraries such as the Bibliothèque nationale de France, the German National Library, and the National Diet Library; research aggregators such as OpenAlex and Crossref; and open-text projects such as Project Gutenberg, Wikisource, and the Directory of Open Access Books. Each adapter runs a controlled set of subject queries against the source’s public API and keeps only records that pass a local relevance filter. The full list, with what each contributed and when it was read, is on the sources page.

Records describing the same work are merged by identifiers — ISBN, LCCN, OCLC, DOI, and source-specific work ids — before falling back to conservative title-and-author matching. A merged record keeps every source it came from, so a book page can show that its description is Harvard’s, its cover Open Library’s, and its PDF link the Library of Congress’s. Where sources disagree on a date or title we keep the variants rather than pick a winner.

On top of the source metadata we add one layer of our own: a taxonomy of 33 traditions. Books are placed by the subject queries that found them and by their recorded subject headings, using a keyword vocabulary per tradition. A book can belong to several. Traditions are meant for browsing, not classification — they answer “what else is like this?”, not “what is this, really?”. Every placement keeps a record of why it was made; the next section explains how.

Content types (monograph, edited volume, thesis, periodical, manuscript, ebook, and so on) and languages are normalized from the source records where possible. Where a source gave nothing usable, the field is left empty rather than guessed.

Since catalog schema 1.18, every one of the roughly 27,000 (book, tradition) placements carries its own inclusion rationale: a recorded basis, a plain-language reason, and the evidence behind it — the controlled subject queries that found the book at its sources and the matching fields in the bibliographic metadata those sources supplied. Nothing is placed by editorial mood; a placement without evidence does not exist.

A placement rests on one of four bases:

Query + metadata match controlled-query-and-metadata

A controlled subject query at a source returned the book, and its retained metadata — subject headings, title, subtitle, or description — also contains the tradition’s vocabulary. Two independent signals; the most common basis by far.

Metadata match metadata-keyword

The book’s bibliographic metadata matches the tradition’s vocabulary, even though the query that found it was aimed at a different tradition. This is how a book found under “occultism” also lands in astrology when its subject headings say so.

Query match controlled-query

A controlled query for the tradition returned the book, but the retained metadata is too thin to corroborate it — common for sparsely cataloged records. The placement stands on the source’s answer to the query alone.

Source classification (retained through merging) retained-source-classification

A source record merged into this book carried the classification, and it was kept when the records were combined. The rarest basis, and always attributed to the source that supplied it.

The evidence is visible wherever a placement is. Each book page has a “Why this book is in the catalog” panel quoting the matched queries and metadata fields, and each tradition page opens with a “How books qualify” breakdown of its placements by basis. The wording is provenance-first on purpose — “matched the controlled Open Library query ‘Occultism’” is a checkable claim; “this is an occult book” is not.

None of this is a judgment about a book’s contents. Membership means relevance, not endorsement: the rationale records why a book is useful to a reader browsing a tradition, not what the book claims or whether anyone here believes it.

This is a broad discovery snapshot, not a complete bibliography. Sources cap their search results, adapters stop at a fixed depth, and the relevance filter is deliberately cautious. Expect gaps — especially for non-Western material, for pre-1800 print, and for anything a national library did not index under a subject heading we queried.

Dates deserve particular care. “First published” is our best reading of the earliest date across the sources we saw, which is often the date of a specific edition, a reprint, or an ebook release rather than the original. Each book page states how confident the date is and where it came from. Do not cite our dates without checking the record behind them.

Descriptions come only from sources — a library abstract, a publisher blurb, a repository summary. We do not write our own, and we do not generate them. Where no source supplied one, the book page says so.

Aggregate figures such as counts per tradition or per language are counts of catalog records, not of works in the world. Refresh cycles re-harvest sources periodically; between refreshes, changes upstream are not reflected here.

A book marked “public” or “read free” is one whose source exposed a full-text link with an access label of public, open, or public domain at harvest time. That label is the source’s, not ours, and it is not legal advice. Rights vary by country; a work in the public domain where a library scanned it may not be where you are. Check the source record and your local law before redistributing anything.

We list direct PDF links only where the source exposed the file itself and reported it as public. Items that are borrow-only, restricted to a reading room, or behind a login are linked to their landing page but never to a file. Where a source explicitly asked not to be hot-linked, we link to its page instead of the file.

Project Gutenberg texts are linked to their gutenberg.org pages, and covers served from Gutenberg follow its published guidance on linking. Cover images from other sources are shown by URL with the source named beneath them; if you own an image and want it removed, see the contact note below.

Catalog metadata is reproduced under each source’s open-data terms, which are recorded on the sources page. Where a source publishes under a specific licence (for instance CC BY 4.0 for the National Bibliography of Scotland, or CC0 for the Biblioteca Nacional de Portugal’s open data) we say so and attribute accordingly.

“Buy” links on book pages point at Amazon. When the site’s Amazon Associates enrollment is active, they may earn the site a small commission if you make a purchase, at no extra cost to you.

Links are derived mechanically from a book’s ISBN: an ISBN-10 is used directly as the product id, an ISBN-13 beginning 978 is converted, and where no usable ISBN exists the link is a search for the title and author. Nothing about a link — its presence, target, or position — is influenced by price, commission, or any commercial relationship. We do not receive or display prices, and we do not track what you buy.

Free-to-read links from libraries and archives are always shown alongside, and above, any purchase link.

Libra Esoterica Pro is an optional paid tier that funds the site’s hosting and harvesting. It adds a research console that can answer questions across the catalog and across the public-domain full texts we have ingested, a semantic search that finds books by idea rather than exact title, cross-book correlation of names, places, symbols, and practices, and exports of research sessions.

The research console is built on a language model. Its answers cite the catalog records and passages they draw on, and you should follow those citations: the model can misread a passage, over-generalize, or state something confidently that the sources do not support. Treat it as a well-read assistant who shows their working, not as a source in its own right.

Everything that is free stays free — browsing, search, ratings, reviews, tags, lists, and shelves. Pro never gates the catalog itself. Plans, prices, and cancellation are on the Pro page.

A contact address has not been published yet — it will appear here as soon as one is set up. In the meantime, please note the catalog id or link of any record you want to write about; it is printed at the foot of every book page.

Rights holders: if a cover image or a linked file should not be here, tell us which record and why, and we will remove the link from the next snapshot and, where we can, immediately.

Type is set in Cormorant (display), Cardo (reading text), and Geist and Geist Mono (interface and identifiers), all served under open licences. Icons are from Lucide.

Catalog data is drawn from the open interfaces of the 36 libraries, archives, and repositories listed on the sources page, and from the open-data work of the communities behind Open Library, Wikidata, Wikisource, Project Gutenberg, OpenAlex, Crossref, DataCite, DOAB, and OAPEN. Language and identifier normalization lean on ISO 639, ISNI, ORCID, GND, and the Library of Congress name authorities.

The mark is a balance drawn as an alchemical sigil — Libra, weighing what the sources say. Thanks to everyone who keeps these catalogs open.