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.
-
Highlight
Color the background of a text selection. Standard use: highlight a key claim worth citing, a numerical result you want in the chapter, or a sentence you intend to quote.
-
Underline
Draw an underline beneath selected text. Useful for marking methods-of-interest, defined terms, or instrument/dataset names that recur across the lit review.
-
Strikethrough
Draw a line through selected text. Use sparingly — in research markup, strikethrough is for claims the paper itself retracts in a later section, or for sentences you disagree with strongly enough to flag.
-
Sticky note
The marginalia primitive. Single positioned note with optional body text — your own marginal questions, links to related papers, or 'TODO: re-read after I finish Bartlett 2019'. No comment-thread replies; FileHop is not Hypothes.is.
-
Text box
Inline typed text on the page. Use for longer marginalia that does not fit a sticky note, or to type a citation reminder next to a passage you want to come back to.
-
Freehand drawing
Stylus or trackpad freehand strokes. The 'reading on a Mac with the Magic Trackpad' primitive — quick scribbles, circles, arrows that do not fit a geometric shape. Strokes are preserved as vector points.
-
Rectangle and circle
Filled or outlined shapes with adjustable stroke width. The figure-circling primitive — outline a panel of a multi-panel figure, box a row in a results table, group a paragraph as 'the methods passage that matters'.
-
Line and arrow
Point at a value, connect two parts of a figure, draw a vector across the page. Useful for marking up a methods schematic, or for visualizing a citation chain across the paper.
-
Stamp
Customizable stamps with template, rotation, and optional top/bottom arc text. The personal-classification primitive — READ / CITE / CITED / Q1 / Q2 / NULL-RESULT / DOI?-NEEDED / METHOD-FOR-CH3. Apply by hand; FileHop does not parse stamps and there is no auto-classification.
-
Whiteout (non-redaction)
Paint a rectangle to hide content visually. NOT destructive — the underlying text is still in the PDF and would be recoverable. For destructive removal of participant identifiers in qualitative transcripts, use the separate Redact tool, not this primitive.
-
Image
Embed a PNG or JPG on the page. Useful for pasting a small reference diagram from another paper, a hand-drawn diagram from your notebook, or a screenshot of a related figure you want side-by-side.
-
Signature
Hand-drawn vector, typed text, or uploaded image. Researchers rarely sign in the reading pass — keep this primitive in mind for signing review-completed forms, co-author sign-off pages, or thesis-committee approval sheets.
-
Text emphasis modes (combined)
Highlight, underline, and strikethrough share an underlying selection model — you can apply them to the same text run in sequence and they layer cleanly on export. Pick a per-color convention (e.g., yellow=claim, green=method, pink=question) and keep it consistent across the lit review for fast skimming later.
“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
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
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
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
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
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
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
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
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? ▼
Can I use this to annotate a manuscript I'm peer-reviewing? ▼
What about Hypothes.is — isn't that the standard for academic annotation? ▼
Can I read on iPad and annotate on the desktop? ▼
Where are annotations stored? Will I lose them if I move the PDF? ▼
Will my marked-up PDF open correctly in Zotero, Acrobat, and Preview? ▼
Can I use this for an embargoed pre-print my colleague sent me? ▼
Can my co-authors and I annotate the same draft at the same time? ▼
Does FileHop do citation extraction or generate a bibliography? ▼
What about Linux? ▼
Is this IRB, HIPAA, GDPR, or FERPA compliant? ▼
I'm a lawyer marking up a contract draft, not a paper. Where do I go? ▼
Citations and further reading
-
Peer-review confidentiality policy
Peer Review Confidentiality (Harvard Medical School, Academic and Research Integrity) → -
Open-annotation context (different job)
Making Peer Review More Transparent with Open Annotation (Hypothes.is) → -
v0.25.0 release line (cited verbatim)
Added in-memory PDF annotation state retention with tap-dot rendering and a replace/new export wizard — FileHop CHANGELOG, 2026-04-17 → -
Reference-manager comparison context
Annotate & Take Notes — Zotero (Harvard Library Research Guide) →
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.