Read with source context
Keep PDFs, metadata review, highlights, source-linked notes, tags and project notes together.
CiteFlow
Read research PDFs, collect notes and keep each quotation close to its source.
Free
Windows 10 1809+ or Windows 11 · x64
Read with source context
Keep PDFs, metadata review, highlights, source-linked notes, tags and project notes together.
Compare evidence
Read sources side by side, keep page-linked notes and review copied passages in Citation History. Reopen the PDF from your library to check the original page.
Search exactly or by meaning
Use exact search without a model or install E5 or BGE-M3 for optional local semantic suggestions.
Before you choose
- Semantic Search is beta and can miss relevant passages or return irrelevant ones; similarity is not certainty.
- OCR, metadata, translation and citation output require review against the source.
- Managed OCR covers eight languages plus script detection; it does not apply or expose page-orientation correction.
Keep one note tied to its PDF page
Use one non-sensitive practice PDF. Review the network disclosure first: source preparation and title translation can make automatic online requests.
- Choose New Project, give it a clear name and select the citation style required for the work.
- Add one copied PDF and let preparation finish. Check the title, authors and year against the PDF instead of accepting automatic details without review.
- Open the PDF from the project library. Choose Note in the viewer toolbar, click beside a passage, write a short comment and select Save. Reopen the page and its note marker to check the connection.
Expected result: a saved note that returns to the page supporting it. If the source details or page are wrong, correct them before using the note or creating a citation.
Follow the research organization guide for more sources, tags, search and checked exports.
See CiteFlow
Select an image for a closer look.
CiteFlow
Read research PDFs, collect notes and keep each quotation close to its source.
Free
Version 1.2.1
Get from Microsoft StoreStandalone EXE
Verify the exact file and view mirrors
- Standalone version
- 1.2.1
- Filename
- CiteFlow-Setup-1.2.1-x64.exe
- Size
- 82,687,958 bytes (78.86 MiB)
- SHA-256
095ce09e409aee50f87b7b71a3f8d837ef15d1bce9ba1bf10395eb8294002ee5- Authenticode
- Not signed
A matching checksum confirms the file matches this release record. It does not prove that software is safe or digitally signed.
More about CiteFlow
Features and working limits
Decision summary
Use a local evidence workspace after sources are collected
Choose exact search for known terms, optional semantic search for related wording and OCR for image-only pages. Reopen the source before relying on any result.
Compare evidence
Read two sources side by side. Citation-aware copy saves passages and citation context in Citation History. Reopen the PDF from the project library to check the original page.
Search exactly or by meaning
Use exact search without a model or install E5 or BGE-M3 for optional local semantic suggestions.
Prepare citations and exports
Run local OCR, then create APA 7, MLA 9, Chicago 18, or IEEE citation drafts to check against the source.
Who this app fits
Good fit when
Choose it when you already have research PDFs and want to read, compare, search, and organize their evidence in one local project.
Choose another tool when
Choose other services for shared cloud editing, database discovery, formal systematic-review screening, protocol reporting or a full writing editor.
Important limits
- Semantic Search is beta and can miss relevant passages or return irrelevant ones; similarity is not certainty.
- OCR, metadata, translation and citation output require review against the source.
- Encrypted, corrupted or poor scans can fail.
- CiteFlow does not automate critical appraisal or guarantee correct research conclusions.
System requirements
- Windows 10 build 17763 or later, or Windows 11; x64 processor.
- Multilingual E5 Large INT8 is about 557 MiB; 4 GiB available RAM required; 8 GB system RAM recommended.
- BGE-M3 Full Precision is about 2.13 GiB; 6 GiB available RAM required; 16 GB system RAM recommended.
- The optional OCR component includes eight text languages plus script detection.
Choose the right search method
Scroll the table horizontally on a narrow screen.
| What you need | Start with | What to check |
|---|---|---|
| Known word, author or identifier | Exact search | PDF text layer or OCR output |
| Related ideas with different wording | Multilingual E5 Large INT8 | About 557 MiB and 8 GB RAM recommended |
| Larger semantic model | BGE-M3 Full Precision | About 2.13 GiB and 16 GB RAM recommended |
| Scanned or image-only PDF | OCR, then exact or semantic search | Selected OCR language; names, numbers and quotations checked against the page image |
What happens from input to result
What is new in 1.2.1
Review source details together, follow optional OCR in the workspace and keep bibliography work more reliably.
- Source-detail warnings appear after automatic preparation. Review multiple sources, open the original PDF and correct missing information.
- Saved source changes remain if you cancel review. Excluding a source affects only the current output; removing it from the project is a separate choice.
- OCR remains optional. Workspace progress distinguishes running, completed, cancelled and failed processing.
- Bibliographies are retained more reliably after citation actions and source review; preparation must finish before generation.
- Corporate-author handling, page-number detection, dialog layout and language switching have improved. Existing projects, citations, notes and settings are preserved during updates.
Turn source PDFs into traceable evidenceKeep every note and claim linked to the page that supports it.
Collect and review sources
Use a documented discovery method, then add the PDFs that belong in the local project.
Read and link evidence
Keep a note or copied passage connected to the source page and compare sources when needed.
Retrieve and verify
Use exact, semantic or OCR-assisted search, then reopen the page before using the evidence.
Installed research tools run without internetReading, notes, comparison, search, citation export and installed OCR run locally. Source preparation and title translation can make automatic online requests; component downloads and update checks are separate.
Stays on your PC
- PDF files, extracted text, indexes and notes remain in the local workspace. Online source preparation can send identifiers or titles, as described below.
- Exact Search runs locally without a semantic model. Semantic Search runs locally after its optional model is installed.
- OCR runs locally after its managed component is installed.
Network actions and external links
- Optional model and OCR installers download only the files you select and check them before use.
- CiteFlow checks for updates at launch if six hours have passed since the previous check.
- The first time Donate opens, CiteFlow requests the public supporter list.
- Source preparation can start automatically when PDFs are imported or a project opens. It can send a DOI or PDF title to metadata registries and request matching publisher records. Manual entry alone does not prevent these background requests.
- When a project needs a translated title, CiteFlow can send it automatically to MyMemory, then Google if the first provider fails.
See the portfolio-level explanation in Network & data.
Local data and network use
- Reading, notes, comparison, search, citation export and installed OCR work without an internet connection.
- PDFs, extracted text, search indexes and notes are stored in the local workspace. Source titles and identifiers can also be sent by the preparation and translation features described below.
- Exact Search runs locally without a semantic model. Semantic Search runs locally after its optional model is installed.
- OCR runs locally after its managed component is installed.
- Update metadata is checked automatically when due, with a six-hour interval; the public Supporters list is requested when Donate first opens.
- Optional model and OCR installers download only the files you select and check them before use.
- Source preparation can start automatically when PDFs are imported or a project opens. It can send a DOI or PDF title to metadata registries and request matching publisher records. Manual entry alone does not prevent these background requests.
- When a project needs a translated title, CiteFlow can send the title automatically to MyMemory, then Google if the first provider fails. Both providers can receive it after a failure.
System requirements
Windows 10 version 1809 (build 17763) or later, or Windows 11; x64 architecture; model-specific memory requirements apply
Frequently asked questions
Does CiteFlow upload my PDF library?
PDF files stay on the PC for reading, notes, search and installed OCR. Source preparation can start automatically on import or when a project opens; it can send a DOI or title to metadata registries and request matching publisher records. Title translation can also run automatically through MyMemory, then Google if needed. Manual entry alone does not prevent these requests. Update checks and the supporter-list request are separate network actions.
Does exact search require a semantic model?
No. Exact search works without E5 or BGE-M3 and depends on the PDF text layer or OCR output.
Is Semantic Search guaranteed to find the right evidence?
No. Semantic Search is beta. It can miss relevant passages or return irrelevant results, and similarity is not proof of a claim.
Which semantic model should I choose?
E5 Large INT8 is about 557 MiB with 8 GB RAM recommended. BGE-M3 Full Precision is about 2.13 GiB with 16 GB RAM recommended.
Can CiteFlow OCR scanned research PDFs?
Yes. The optional managed OCR component supports eight text languages plus script detection, but names, numbers, tables and citations must be checked against the page image.
Which citation styles are included?
The app supports APA 7, MLA 9, Chicago 18 author-date, Chicago 18 notes-and-bibliography and IEEE drafts.
Does CiteFlow replace a systematic review platform?
No. It supports local evidence organization after collection and does not replace database searching, protocol registration, formal screening, deduplication or critical appraisal.