Skip to main content

Annotate research papers offline — for literature review, peer review, and embargoed work

Highlight, marginalia, freehand on figures, and stamps as a personal classification system. Your 60 papers stay on your Mac or Windows machine; nothing uploads. Works alongside Zotero, Mendeley, or whatever reference manager you already use.

The moment you reach for this guide

You are three weeks into a lit review on 60 papers, or you have a peer-review assignment due next week, or a colleague has sent you an embargoed pre-print and asked for marginalia. Your reference manager's PDF reader handles highlights but you want more — sticky notes for marginal questions, freehand circles around figures, stamps for READ / CITE / Q1 / NULL-RESULT as a personal classification system. You also do not want to upload the embargoed pre-print or the peer-review draft to a cloud annotator. This guide walks the offline workflow on a desktop app.

Where FileHop fits in your existing stack

FileHop is NOT a reference manager. Zotero, Mendeley, Papers, Citavi, Paperpile, and EndNote continue to do the library, collection, metadata-harvesting, and citation-export jobs they were built for. FileHop sits below that layer — the annotation pass on the PDF itself happens here, and the paper is catalogued upstream in the reference manager. The marked-up PDF (flattened on export) renders correctly in Zotero's built-in PDF reader, Adobe Acrobat, and macOS Preview, so the file goes back into your library unchanged.

Hypothes.is is the right answer for open, social, public annotation of web-published research — post-publication discussion, voluntary open peer review, collaborative pedagogy in the classroom. It is not the right answer for confidential peer review of a submitted manuscript or for an embargoed pre-print, because the reviewer's confidentiality obligation prohibits uploading the manuscript to a third-party service. FileHop handles the offline-confidential job; Hypothes.is handles the open-public job. Two different jobs, two different tools.

The 13 annotation primitives, in research-workflow terms

Same primitives the lawyer-cluster sibling uses for contract review, relabeled for the lit-review job. Map your reading intent to the right primitive before you start — it makes the marked-up paper easier to scan months later when you come back for the thesis chapter.

From the v0.25.0 changelog (April 17, 2026)
“Added in-memory PDF annotation state retention with tap-dot rendering and a replace/new export wizard”

Two practical consequences for the reading pass: (1) you can close the PDF and reopen it without losing marks-in-progress — annotation state is held in memory and rendered with tap-dot markers as you place them; (2) when you export the marked-up PDF, the wizard asks whether to overwrite the source or create a new flattened file. The default is 'create a new file', which keeps the source PDF clean for your reference-manager library.

Workflow A — Single-reviewer lit-review pass

The mode you'll be in 90% of the time. Reading one paper at a time, annotating as you go, stamping at the end. Annotations live as sidecar JSON next to the source PDF; the source PDF itself is unchanged. Export to a flattened marked-up PDF when you're ready to push it back to your reference manager or send a snippet to a co-author.

  1. 1

    Open the paper in FileHop

    Drag the PDF from your reference-manager library folder (Zotero's storage folder, Mendeley's Watched Folder, your manual filesystem under /papers/2026/) into the FileHop desktop app. The source PDF is read in place — no upload, no account required for annotation. The sidecar JSON that holds your markup is written beside the PDF in the same directory. If you keep your papers on an encrypted volume (FileVault on macOS, BitLocker on Windows, VeraCrypt for either), the sidecar lives there too.

    See the full researcher workflow set →
  2. 2

    Read and mark as you go

    Highlight claims worth citing, underline methods, sticky-note your marginal questions, circle figures with rectangle or freehand, draw arrows across citation chains. The toolbar carries all 13 primitives. Color is a personal convention — pick a scheme (e.g., yellow=claim, green=method, pink=question, blue=quote-verbatim) and keep it consistent across the lit review. Annotations are reversible while the paper is still in FileHop's editable layer: move, edit, recolor, or delete anything you've placed.

  3. 3

    Stamp the paper with your classification at the end of the read

    Drop a stamp on the title page when you finish: READ + the date + one of CITE / CITED / Q1 / Q2 / NULL-RESULT / METHOD-FOR-CH3 / DOI?-NEEDED. Stamps are entirely your convention — FileHop does not parse them, there is no auto-classifier, no machine-learning tag. The point is to make the lit-review folder skimmable when you come back to it three months later for the chapter.

  4. 4

    Export to a flattened marked-up PDF (optional)

    Use the export wizard (v0.25.0 — the wording is 'replace / new export'). The default is to create a new file ending in _annotated.pdf, leaving the source clean. The marked-up PDF flattens annotations into PDF page content, so it renders correctly in Zotero's built-in reader, Adobe Acrobat, macOS Preview, Foxit, and any other PDF reader — your co-author does not need FileHop installed to see your marginalia. Keep the unflattened original PDF and its sidecar JSON if you want to keep editing later; archive both side-by-side if you're done with the paper.

Workflow B — Multi-reviewer consolidation (co-authors, peer-review panels)

When several people need to mark up the same manuscript without uploading it to a shared cloud workspace. Each reviewer annotates their own copy on their own machine and flattens; one person merges the flattened marked-up PDFs into a single review packet. The editorial wedge: the SERP does not walk this asynchronous-merge pattern, but it is the right answer for confidential peer review (where the journal's confidentiality clause prohibits uploads) and for embargoed-material co-author review.

  1. 1

    Distribute the manuscript to each reviewer

    Send each reviewer their own copy of the same source PDF — encrypted attachment, secure file-share inside your institution, or a USB drop if the confidentiality obligation is strict. Each reviewer keeps their copy local. They do NOT share-link via a cloud annotator. There is no shared file and no concurrent edit — that's the trade against cloud collab, and it's the trade you wanted for this draft.

  2. 2

    Each reviewer marks up their own copy

    Reviewer 1 reads and annotates in FileHop on their machine. Reviewer 2 does the same on theirs. Reviewer 3, same. Each reviewer is using the 13 primitives independently — there is no live cursor presence, no comment threads, no shared workspace. Convention: ask each reviewer to use a distinct annotation color (Reviewer A in red, Reviewer B in green, Reviewer C in blue, lead reviewer in orange) and to include initials in the filename of their flattened export so attribution is obvious at a glance. This convention substitutes for the per-comment attribution that cloud-collab tools provide automatically.

  3. 3

    Each reviewer flattens and sends the marked-up PDF back

    Each reviewer exports a flattened _annotated.pdf via the export wizard (Workflow A, Step 4). They send the flattened PDF to the lead reviewer (or to the article's first author) via the same secure channel the source was distributed on. The flattened PDF renders identically in any PDF reader — the lead reviewer does not need FileHop to see the markup, which matters when the lead is on a managed institutional machine where they cannot install new software freely.

    Open the Merge PDF tool →
  4. 4

    Lead reviewer merges the flattened PDFs into a single review packet

    Open FileHop's Merge PDF tool and combine the flattened marked-up versions in reviewer order — typically the clean draft first, then Reviewer A markup, Reviewer B markup, Reviewer C markup, then your own consolidated markup last. Page numbers run continuously through the packet, and you can scroll to see every reviewer's annotations on each section side-by-side. The merge operation runs locally; the PDFs do not leave your machine for the merge step. Distribute the packet to the article author or the panel chair via the same secure channel.

⚠ Honest limit

This pattern is for small rounds — for heavy synchronous review, look at cloud collab

Workflow B works well for 2–5 reviewers across one or two rounds. For a manuscript that's going to bounce back and forth across a dozen reviewers in synchronous editing sessions, or for a large open-peer-review pilot where transparency is the explicit goal, a cloud workspace (Acrobat Share for Review, an institutional collaboration platform, or Hypothes.is for the open-peer-review case) is the right answer — and that's a different category of software with a different privacy posture you'd evaluate separately for the confidential-material question.

What this workflow is NOT

Read this before you switch. The boundary is the wedge — knowing what offline annotation cannot do is what makes the right-job mapping work. These are not caveats; they are explicit scope.

  • NOT a reference manager. Zotero, Mendeley, Papers, Citavi, Paperpile, and EndNote stay in their lane — library, collection, metadata harvesting, citation export, shared-library sync. FileHop annotates one PDF at a time. The two layers stack; FileHop does not replace the layer above it.
  • NO citation extraction, BibTeX / RIS export, or DOI lookup. Use your reference manager for that. If you need a quick BibTeX entry while reading, copy the DOI from the PDF and paste it into Zotero or Mendeley — that round-trip is the right one.
  • NO AI literature synthesis or smart cross-paper summary. Elicit, Scite, SciSpace, Consensus, and ResearchRabbit are the tools for that job. FileHop annotates; they synthesize. The article must not blur the line — those tools take uploads, which is the posture you may be specifically avoiding for confidential material.
  • NO real-time co-annotation, comment-thread replies, @mentions, or live cursors. Workflow B is the asynchronous offline alternative. If you need synchronous markup, Acrobat Share for Review or a cloud workspace is the right answer — and you'd evaluate that vendor's privacy posture for this specific manuscript.
  • NO IRB, HIPAA, GDPR, or FERPA certification. We do not make compliance claims. The architectural choice is 'no upload by default for the annotation workflow', which is a different and verifiable statement than 'regulatory-compliant'. Your institution's compliance officer judges your workflow against your policy; the local-processing piece is what this tool gives you.
  • NO iPad, iOS, Android, Linux, or web build. Mac and Windows desktop only. If you read on iPad, GoodNotes / Notability / LiquidText / MarginNote / Highlights / PDF Expert handle the reading pass well — route the marked-up PDF back to the desktop for export/merge/stamping when you're done.
  • Annotations are non-destructive sidecar JSON. They live in a JSON file next to the source PDF. If you move the PDF without the sidecar, the annotations are lost. Either keep the JSON beside the PDF as a pair, or flatten to a marked-up PDF as soon as the reading pass is done so the marks travel with the file.
  • NO PDF-vs-PDF document compare. If you need to diff manuscript v3 against manuscript v2, Acrobat Pro Compare Files or a specialist tool handles that. FileHop annotates one PDF at a time.

How offline annotation compares to the alternatives

Workflow comparison only, not a complete product comparison. Many researchers run a reference manager AND a desktop annotation tool AND an iPad reader — each for a different step of the reading pass. Pick the tool that matches the privacy posture of the specific paper or manuscript in front of you.

Workflow capability FileHop (offline desktop) Reference-manager readers (Zotero, Mendeley) iPad apps (MarginNote, LiquidText, Highlights) Cloud collab (Hypothes.is, Acrobat Share, Smallpdf)
Where files live Local-only; no upload; sidecar JSON next to PDF Local + opt-in cloud sync Local + iCloud or app-specific cloud Cloud; files uploaded for the session
Annotation primitives 13 (highlight, underline, strikethrough, rectangle, circle, line, arrow, freehand, sticky, text box, whiteout, image, stamp, signature) 3–5 (highlight, area-select, sticky note, sometimes ink) 10–15 depending on app; strong stylus / ink support Variable; usually highlight + comment + drawing
Multi-reviewer pattern Asynchronous offline merge (Workflow B) Shared library with synced annotations (opt-in cloud) Single-machine app; awkward multi-reviewer Real-time co-annotation with sign-in
Export portability Flatten to standard PDF — renders in Acrobat, Preview, Zotero, any reader Annotations stay inside the manager's database; export is workable but app-bound Often locked to proprietary format; PDF export can degrade or omit infinite-canvas elements PDF export with cloud-flavored comments
Reference-manager integration Adjacent, not built-in — flat PDFs go back to the manager unchanged Native — IS the manager Some integration (e.g., Zotero plugins for MarginNote) None — outside the reference-manager workflow
OS support Mac + Windows desktop only Mac / Windows / Linux / web / iOS / Android iPad primary; some Mac builds Web (any modern browser)
Real-time co-annotation No — asynchronous only No (annotation conflicts get merged later) No Yes (Hypothes.is groups, Acrobat Share)
Citation extraction / BibTeX No Yes (RIS / BibTeX / DOI lookup) Limited (some MarginNote outline export) No
Peer-review-confidentiality fit Strong (architectural — no upload) Workable if cloud sync is OFF Workable for the reading pass; export awkward Weak — files uploaded to third party
Embargoed-material fit Strong Workable with cloud sync OFF Workable Weak
Cost Free desktop install (no subscription for annotation) Free (Zotero); freemium (Mendeley) Paid (one-time or subscription) Freemium / subscription

What 'offline' actually means here

Plain statement of what stays on your machine and what doesn't. No marketing language — just the architecture.

  • Annotations are saved to a sidecar JSON file next to the source PDF. Move the PDF; move the JSON. Or flatten to a marked-up PDF and archive both.
  • The source PDF is never modified in the annotation pass. Flattening creates a new file by default (export wizard from v0.25.0). If you explicitly choose to overwrite the source, the app respects that choice.
  • No telemetry on the content you're annotating. FileHop does not send the PDF, the annotations, the highlights, or the marginalia to a server.
  • No account required for annotation. No sign-in, no email required for the annotation workflow itself.
  • Mac and Windows desktop only. Linux, iPad, iOS, Android, and web are out of scope — credit the iPad apps for the iPad reading pass and route export/merge to the desktop.
  • Your local backup hygiene is still your responsibility. Time Machine, OneDrive backup of your local files, an encrypted external drive — all standard. The 'no upload by default' architecture removes one class of leak risk; it does not replace your backup discipline.

FAQs

Is this a replacement for Zotero, Mendeley, Papers, Citavi, or Paperpile?
No. FileHop annotates one PDF at a time on the desktop. Your reference manager handles the library, the citation export, the DOI lookup, and the metadata harvesting. The marked-up PDF (flattened on export) goes back into your reference-manager library unchanged — Zotero's built-in reader, Adobe Acrobat, and macOS Preview all render the flattened annotations correctly. The two layers stack; one does not replace the other.
Can I use this to annotate a manuscript I'm peer-reviewing?
Yes — and this is the workflow many reviewers reach for when the journal's confidentiality clause prohibits uploading the submitted manuscript to a third-party service. Harvard Medical School's research-integrity statement, for example, is explicit that 'integrity and confidentiality within the peer review process is mandatory'. The Nature, Cell, NEJM, and NSF / NIH panel-review policies are analogous. Offline annotation removes the third-party-server class of confidentiality risk by design. It is not a regulatory certification; it is an architectural choice.
What about Hypothes.is — isn't that the standard for academic annotation?
Hypothes.is is the right answer for open, social, public annotation of web-published research — post-publication discussion, voluntary open peer review, collaborative pedagogy. It is not the right answer for confidential peer review of a submitted manuscript or for an embargoed pre-print, because the reviewer's confidentiality obligation prohibits uploading the manuscript to a third-party service. Two jobs, two tools — credit Hypothes.is for what it does well, and use the offline workflow for the confidential case.
Can I read on iPad and annotate on the desktop?
Yes — and many researchers do exactly this. iPad apps (GoodNotes, Notability, MarginNote, LiquidText, Highlights, PDF Expert) handle the stylus-driven reading pass well. When you're ready to finalize the markup, export the marked-up PDF from the iPad app and open it in FileHop on the Mac or Windows desktop for the merge, flatten, or stamping pass. Use the right tool for the right step rather than expecting one app to cover the whole workflow.
Where are annotations stored? Will I lose them if I move the PDF?
Annotations are saved as a sidecar JSON file next to the source PDF. If you move the PDF, move the JSON file too — they live as a pair. The safer alternative is to flatten to a marked-up PDF as soon as the reading pass is done (export wizard, v0.25.0). The flattened file embeds the annotations as PDF page content, so it travels alone and renders correctly in any PDF reader.
Will my marked-up PDF open correctly in Zotero, Acrobat, and Preview?
Yes. Flattening converts annotations to standard PDF page content (or to standard PDF annotation objects, depending on the primitive). Zotero's built-in PDF reader, Adobe Acrobat, macOS Preview, and PDF Expert all render the flattened output. Sticky notes appear as PDF text annotations; highlights as PDF highlight annotations; freehand strokes and shapes as flattened drawings. Your co-author does not need FileHop installed to see your markup.
Can I use this for an embargoed pre-print my colleague sent me?
Yes — this is one of the high-trust use cases. Embargoed pre-prints are exactly the material you want to keep off cloud annotators. The offline workflow keeps the PDF on your machine; the sidecar JSON stays beside it; the flattened export, when you make one, stays in your local file system unless you explicitly send it. The architectural posture matches the embargo requirement.
Can my co-authors and I annotate the same draft at the same time?
Not in real time. FileHop is asynchronous and single-machine. The pattern is Workflow B above: each co-author annotates their own copy on their own machine, flattens, and sends the marked-up PDF back to the lead reviewer, who merges the flattened PDFs into a single review packet with the Merge PDF tool. If you need synchronous markup, the right tool for that job is Acrobat Share for Review or a comparable cloud workspace — and you would evaluate the vendor's privacy posture for the specific draft.
Does FileHop do citation extraction or generate a bibliography?
No. Citation extraction, BibTeX / RIS / DOI export, and bibliography generation are reference-manager jobs — Zotero, Mendeley, Papers, EndNote, Paperpile, Citavi. FileHop annotates the PDF itself. The two layers stack: the paper is catalogued upstream in the reference manager, and the annotation pass on the PDF happens here.
What about Linux?
Mac and Windows desktop only. Linux researchers (common in HPC and STEM environments) typically reach for Okular, Xournal++, or a qpdf-and-stylus pipeline. FileHop does not currently build for Linux; we surface this constraint up front rather than pretending the workflow covers every operating system.
Is this IRB, HIPAA, GDPR, or FERPA compliant?
FileHop does not carry IRB, HIPAA, GDPR, or FERPA certification — we do not make compliance claims. The architectural choice is 'no upload by default for the annotation workflow', which is a different and verifiable statement than 'regulatory-compliant'. Your institution's compliance officer judges your workflow against your policy; the local-processing piece is the part of the workflow this tool gives you. The certification piece is upstream of any tool.
I'm a lawyer marking up a contract draft, not a paper. Where do I go?
Same underlying technology, different workflow framing: the lawyer-cluster sibling guide walks the contract-review and partner-review redline-consolidation pattern. The 13 annotation primitives, the sidecar JSON storage, the flatten-to-PDF export, and the multi-reviewer merge pattern are all shared between the two articles. The lawyer guide swaps lit-review framing for contract-review framing.

Annotate your next paper on your own machine

Free desktop install. No account required for annotation. The file stays on your machine. Mac and Windows only. If you do this kind of reading work regularly — annotate papers, anonymize manuscripts, OCR scanned archival material, redact qualitative transcripts — the researcher persona page walks the broader workflow set, and the related guides below cover adjacent steps.