Papyrus keeps the whole research workflow beside the paper.

One Library

A paper should not fragment into separate tools. The PDF, metadata, reading progress, tags, notes, project role, evidence, and citation key all belong to the same object. Papyrus keeps that object coherent from first import through the final export.

The example throughout this page follows a single Nature article: A 5.3-million-year-old deep-sea whale necropolis in the Diamantina Zone. The same record appears in the library, opens in the reader, carries Markdown notes, files those notes onto a notebook, supports a project archive, and exports through standard citation formats. The figures here show the Mac; iPad and iPhone have pages of their own.

Capture & Organize

Import should produce a usable record, not a cleanup chore. Papyrus accepts PDFs, folders on Mac, DOI and arXiv identifiers, BibTeX, and RIS. It can read embedded PDF metadata, fetch scholarly metadata, and keep field-level provenance so you can see where a title, author list, year, DOI, or abstract came from when sources disagree. After a library-wide refresh, the summary lists which papers changed and which fields moved.

Scanned papers used to arrive empty. When an opening page carries no extractable text — a photocopied chapter, a raster-only scan — Papyrus reads it with bounded on-device OCR and recovers the title, byline, identifiers, year, and abstract from the image itself. Nothing is uploaded to do it, and OCR-derived fields are marked as such and ranked below publisher metadata, so a reading of the page never outranks a source. An unresolved DOI no longer stops the lookup either: Papyrus falls through to the paper's arXiv, PubMed, or ISBN identifier instead of giving up on the first one that fails.

A refresh that changes nothing now says why. Already current, no identifier we can look up, offline, a field you locked against a conflicting source — each is reported as itself rather than as a uniform “Metadata Refreshed”, and a run over many papers shows its progress and can be stopped partway.

Safari can hand papers over directly. The bundled Safari extension captures the paper on the current tab — with its PDF when one is available — or saves any PDF, DOI, or arXiv link from the right-click (Mac) or long-press (iPad) menu. It captures from inside your browser session, so a paper behind a publisher sign-in you already have access to saves too, not just open-access files. The popup shows exactly what was captured and offers folder, tag, and project pickers so a paper lands organized instead of joining an inbox pile.

Organization has several layers because research changes shape over time. Tags describe topics. Folders keep stable archives, and a paper can live in more than one. Flags give seven colors you can name yourself, each with its own shape so they stay tellable apart at a glance. Smart Collections stay current as saved filters — several values per rule, matched on Any or All. Projects gather papers for a draft, review, or literature archive without forcing the global library to mirror a manuscript outline.

Read & Annotate

Reading happens in a native window, not a web viewer embedded in a database. On Mac and iPad you can highlight, underline, strike through, draw freehand ink, leave notes, and set bookmarks. The reader remembers progress and keeps annotations attached to the paper instead of scattering them into a separate notebook.³ Where you stopped, how far in you are, and every page you bookmarked also surface on the paper itself, in the Reading card.

On iPad and iPhone the page can be laid out two ways at once. A Layout Options panel sets the page layout — one column or two-up, with the first page shown alone so spreads line up the way a bound book does — and can split the reader into two panes, side by side or stacked, each holding the same document at a different place. Comparing a figure on page 3 against the discussion on page 14 stops being a scrolling exercise. Search highlights fan out to both panes, and navigation follows whichever pane you touched last.

Figure-heavy articles need more than a citation key. Papyrus lets the article page, the figure, the reading note, and the eventual claim remain adjacent. A Markdown notes pane renders headings, lists, tables, code, and LaTeX math, so a figure comment or draft-ready interpretation stays beside the source that prompted it.

Notebooks

Notes are written one paper at a time, but they are needed by theme. A notebook is a shelf for a topic, a chapter you are drafting, or a reading group. Filing a note onto one copies nothing and moves nothing: the note still belongs to its paper, still says which paper that is, and one click takes you there — with a way back to the shelf. A note can sit on several shelves at once, so “Optics” and “Chapter 3” never have to fight over it.

Add Notes… fills a new shelf from what you have already written, with a search field and an unfiled-only filter, so years of notes take moments rather than a note-by-note pass. A shelf can be searched by a note's text or by the paper behind it, and ordered four ways: recently edited, grouped by paper, by title, or oldest paper first — which reads your own notes forward through the literature, the way a related-work section wants to be written. Select several notes to move, unfile, or trash them together. All Notes lists everything you have written, filed or not, from your very first note.

A whole shelf exports as one Markdown document that matches what is on screen — the current order, the current search — and carries a papyrus:// link so Hookmark, Obsidian, or a plain text plan can point straight back at it.

Projects & Evidence

Projects turn a library into a working argument. A paper can belong to several projects, and each project can give that paper a different role. The Nature article can be an anchor source in a Diamantina whale-fall archive, while adjacent papers provide ecological context, dating methods, or organism-specific background.

The Evidence Board is the place where Papyrus goes beyond storing citations. Each evidence item records the claim you are making, the quote or observation that backs it, your interpretation, the page, the citation key, a section role, and a confidence score. A project starts to read like an argument before the manuscript exists.

Search, Sync & Privacy

A research desk has to be searchable and inspectable. Papyrus indexes titles, metadata, notes, annotations, and extracted PDF text so a phrase can lead back to the paper, page, and context where it was useful. The reading queue keeps active work visible, while filters and Smart Collections surface papers by tag, journal, year range, favorite status, flag, document type, or PDF availability.

Local-first is the default, not an add-on. Your library, PDFs, notes, annotations, search index, and project evidence live on your device. There is no Papyrus account and no telemetry. Network access is limited to actions you choose, such as metadata lookup or sync. iCloud Sync uses your own iCloud account to keep a library current across Mac, iPad, and iPhone, while WebDAV sync remains available for a server destination you control and dated restore points.⁴

Export & Citation

The exit path is as important as the import path. A single paper, a notebook, or a whole project can leave as BibTeX, RIS, CSL JSON, Markdown, or a full bundle. Project export can include papers, notes, evidence, citations, and a report that flags missing PDFs, duplicate cite keys, sparse metadata, or unresolved citation problems before you hand work to another tool.

A paper can also be linked to, not just exported. Every paper and every notebook has a permanent papyrus:// address — copyable as a plain link or as Markdown, and anchorable to the page you are on. Because the identifier is the paper's sync identity, the link survives renaming, re-tagging, moving between folders, trashing and restoring, even replacing the PDF, and it resolves on every device signed in to your library.

That is what makes Hookmark work properly: hook a paper to the manuscript section, the task, or the email that sent you to it, and come back to it later on the page you left. Papyrus also answers get-current-item over x-callback-url, so Hookmark, Shortcuts, Alfred, Keyboard Maestro, or Drafts can ask what you are reading right now and get back the title, the link, and the PDF's path. Read the full URL scheme →

Citations can also be copied in common styles without leaving the library. The reference for peng2026whaleNecropolis, formatted six ways:

@article{peng2026whaleNecropolis,
  title   = {A 5.3-million-year-old deep-sea whale necropolis in the Diamantina Zone},
  author  = {Peng, Xiaotong and Zhou, Peng and Song, Xikun and Bianucci, Giovanni and others},
  journal = {Nature},
  doi     = {10.1038/s41586-026-10546-z},
  year    = {2026}
}