A library of the ancient world that lives on your own computer. Nabu downloads freely licensed ancient texts — Homer, the Bible, cuneiform tablets, the Chinese classics and far more — and lets you search them all at once, offline. It is made for scholars, students and anyone who reads old languages: look up a word such as μῆνις, the first word of the Iliad, and every passage that uses it comes back with an exact citation and the license it is held under.

Try it: the Quickstart takes you from download to your first search in about ten minutes.

Nabu is a piece of personal research infrastructure that gathers the world’s openly licensed digital corpora of antiquity — Homer and the Greek canon, the Latin classics, the documentary papyri of Egypt, the Latin inscriptions of the Roman empire, the Sanskrit tradition and the Pali canon, cuneiform tablets, three millennia of Egyptian sentences, the Masoretic Hebrew Bible with its Targums and Peshitta, the Hittite tablet corpus, the universal CDLI catalog of cuneiform, the Ugaritic tablets, a millennium of Syriac, the inscriptions of pre-Roman Italy and Sicily, the New Testament in up to fifteen parallel witnesses, the complete corpus of Old English poetry and the first printed books in English, the classical Chinese library with the Buddhist canon, the Korean dynastic records, the Islamicate library in Arabic and Persian, the Tibetan Buddhist canon, the inscriptions of Southeast Asia and the palm-leaf lontars of Bali, the stone record of South India with the classical Tamil canon and its commentators, the Gurmukhī scripture of the Sikh canon, the Middle Iranian, Tocharian and Turkic manuscripts of the Silk Road, the medieval lyric of Galicia, the Japanese classical and public-domain libraries, the Slavic shelves from Old Church Slavonic to the Balkan oral epic, the runestones of Scandinavia, and the colonial record of the indigenous Americas — into a single library on the scholar’s own disk. Everything is stored as plain files plus SQLite: it is searchable by word or by dictionary lemma, citable to the exact verse or tablet line, explicit about every text’s license, and rebuildable from its canonical sources at any time. Because the library also exposes a read-only Model Context Protocol server, the AI assistants a researcher already uses can search, quote, and cite the whole collection while remaining structurally unable to alter a letter of it.

The project takes its name from the Mesopotamian god of scribes, patron of the tablet house and divine custodian of Ashurbanipal’s library. It is neither a website nor a reading application: it is a pipeline and a database, operated from the command line, and designed to outlive the services it draws from.

Find your desk

The collection is organized as 30 research desks — scholarly hats, each gathering the shelves, instruments and terminal setup for one tradition. Find yours:

• The Classicist — Greek and Latin letters read whole, Homer to the late grammarians.
• The Romanist — the Latin-to-vernacular continuum, charter Latin to Roland to the troubadours.
• The Papyrologist-Epigraphist — reads what survives on stone, sherd, papyrus and tablet, lacunae and all.
• The Slavicist — Cyril and Methodius to the damaskini, canon to vernacular.
• The Germanicist — Gothic and Old English to the Norse sagas and the runestones, the word-hoard of all three branches.
• The Celticist — from Lepontic stones to the Old Irish glossators.
• The Italicist — the languages of pre-Roman Italy, Oscan to Etruscan to Raetic.
• The Comparative Indo-Europeanist — laryngeals, reflex chains, the long descent of words.
• The Biblical scholar — one text across Hebrew, Greek, Latin, Syriac, Coptic and English witnesses.
• The Hebraist — Masoretic vowels, Qumran consonants, the Aramaic of the Targums.
• The Syriacist — the Peshitta and the estrangela bookshelf.
• The Ethiopicist — Aksum to the scriptoria: Enoch, Jubilees, and the Geʿez Bible.
• The Arabist — the Islamicate library whole, Quran and hadith to falsafa and adab.
• The Hittitologist — Anatolia in cuneiform, KBo and KUB by tablet and line.
• The Assyriologist — Sumerian, Akkadian, Ugaritic, Hittite: the tablet world entire.
• The Egyptologist — hieroglyphs to Coptic, one language across four millennia of script.
• The Iranologist — the Avesta to the Achaemenid inscriptions, Old Iranian liturgy toward Middle Persian.
• The Indologist — Veda to sastra, the Sanskrit library and its instruments.
• The Buddhologist — the dharma across the Pali, Sanskrit, Chinese and Tibetan canons.
• The Tibetologist — the Land of Snows from Dunhuang documents to the complete Derge canon.
• The Koreanist — the hanmun state record and the first hangul vernacular.
• The Southeast Asianist — the Indic cosmopolis at its eastern edge: Sanskrit charters beside the vernaculars they seeded.
• The Sinologist — the classical Chinese written world and its phonological deep past.
• The Japanologist — Old Japanese song to the Sino-Japanese dictionary tradition.
• The Turkologist — the runiform steppe to the Chagatai chancery, one literary continuum.
• The Mongolist — the Secret History’s empire in its own words, hanzi transcription beside romanization.
• The Americanist — the colonial record of the indigenous Americas, Zapotec testaments to the Andean chronicle.
• The Africanist — Ajami manuscripts beyond the Nile–Ethiopia axis, Hausa praise poems to Wolof Murid verse.
• The Balticist — the Lutheran-print littoral, Mancelius’ Latvian to Agricola’s Finnish.
• The Librarian — the owner’s own shelves: dossiers, library, notes, and the sources’ own records.

Each desk is a reader’s-eye view of the same collection; the full directory, with every desk’s member shelves and recipes, is Research desks.

The holdings, in brief

As of 7 November 2026, the catalog records 2,854,118 documents comprising 128,499,000 passages in 277 language codes across 327 live sources — from proto-cuneiform tablets of the late fourth millennium BCE to Meiji-era Japanese. Classical Arabic is the largest language on the shelves (33.3 million passages), ahead of Early Modern English (28.8 million) and Literary Chinese (20.7 million). Beneath every shelf runs the lect layer: historical-stage ladders, stage-aware search, and honest reconstruction display through the nabu-lects registry. Together with 5,281,593 dictionary entries across 956 reference shelves and 25.5 million gold-standard lemma annotations in 53 languages (a further 36.2 million machine-suggested annotations are carried at an honestly labelled silver tier). All figures on this site are read from the live catalog, never estimated, and carry the date on which they were read.

A survey of the collections is given on The Library page; the full attribution and licensing record is on Sources.

Recently on the shelves

The latest entries from the News page, where every wave is told and dated:

• 7 November 2026 Early-modern Europe, the Latin liturgy, and three new ways to search — Four November days: the early-modern European print wave from the Stationers' Register to the mazarinades, six liturgy and chant sources with a chant-identity layer, East and Inner Asia breadth from the Tangut rhyme dictionary to the Qing Pentaglot — and three new search surfaces that open the library's annotation layers, scholars' token glosses, and upstream morphology to direct query.
• 3 November 2026 From the Suda to the South Seas — seven text sources and twenty-six dictionaries — One development window: the complete English Suda, the Hill Museum's Syriac and Ethiopic manuscript catalogue, the Patristic Text Archive, the Sixth Council Pāḷi canon with its printed apparatus, 312 Ming examination records, the library's first Oceanic running text — and a twenty-six-shelf dictionary wave from Old Swedish law glossaries to the Digital Pāḷi Dictionary.
• 2 November 2026 The biggest single day — fifteen new sources, from Ganjoor to the Patrologia Graeca — Two development windows in one day: the demo image went multi-arch with its release rail, a silver scansion lane and the Elizabeth Bible landed, and then fifteen sources arrived in one wave — the classical Persian poetry of Ganjoor under a fresh grant, 33 volumes of Byzantine Greek, 99,571 pages of Tibetan transcriptions, four biblical corpora and three new dictionaries. The library passed 121.5 million passages.

A single verse, many witnesses

One illustration, pasted from a live run (12 July 2026; trimmed lines are marked). A single Gospel citation rendered across the aligned witnesses — Greek, Latin, Gothic, Old Church Slavonic, Old English, and more — each carrying its license label. Since this run, the Sahidic and Bohairic Coptic New Testaments have joined the hub (13 July 2026), bringing the registered witnesses to fifteen:

$ bin/nabu align "MARK 2.3"
MARK 2.3 — New Testament (parallel witnesses)
  13 of 13 witnesses attest this ref

greek-nt — The Greek New Testament [grc]   license: nc
  urn:nabu:proiel:greek-nt:6563
    καὶ ἔρχονται φέροντες πρὸς αὐτὸν παραλυτικὸν αἰρόμενον ὑπὸ τεσσάρων.

latin-nt — Jerome's Vulgate [lat]   license: nc
  urn:nabu:proiel:latin-nt:10368
    et venerunt ferentes ad eum paralyticum qui a quattuor portabatur

gothic-nt — The Gothic Bible [got]   license: nc
  urn:nabu:proiel:gothic-nt:37435
    jah qemun at imma usliþan bairandans, hafanana fram fidworim.

marianus — Codex Marianus [chu]   license: nc
  urn:nabu:proiel:marianus:36421
    Ꙇ придѫ къ немоу носѧште ослабленъ жилами. носимъ четꙑрьми.

wscp — West-Saxon Gospels [ang]   license: nc
  urn:nabu:proiel:wscp:102359
    & hi comon anne laman to him berende, þone feower men bæron.

WEB (English) — Mark [eng]   license: open
  urn:nabu:eng-web:mrk:2.3
    Four people came, carrying a paralytic to him.

… (Armenian, SBLGNT, Clementine Vulgate, and the four CCMH OCS witnesses trimmed)

Every result is a resolvable citation — a stable URN pointing into the local catalog — rather than an unverifiable quotation. The same discipline governs all of the tools: dictionary citations resolve to live passages, intertext hits name the shared phrase, and every surface carries its license class.

Principles

• License honesty. Every ingested text keeps its upstream license, recorded per document and displayed on every surface. Roughly 89% of documents are public-domain or attribution-class; non-commercial and no-derivatives materials are handled under correspondingly stricter rules. See Sources.
• Longevity over convenience. Upstream projects restructure, lose funding, and disappear. The canonical layer is plain files under git; all databases are derived and rebuildable; texts an upstream deletes are retained, honestly labelled. Storage is deliberately boring: files, git, SQLite.
• Citations, not summaries. The unit of work is the citable passage. Search results, dictionary entries, alignment rows, and machine-readable exports all carry stable URNs.
• Measured claims. Corpus numbers are snapshots of one live installation, dated where they appear.

Where to go next

• The Library — what is on the shelves: collections, periods, sizes, licenses.
• Tools — the command-line instruments, organized by scholarly task.
• Examples — worked walk-throughs for a classicist, a Latinist, a papyrologist, a sinologist, a slavist, a comparativist, an Old English scholar, an assyriologist, a hittitologist, a biblical scholar, an egyptologist, a student of pre-Roman Italy, a Hebraist, and a southeast asianist.
• About — what this project is, who it is for, and how it is built.