PaperQuay is a literature manager designed for students, researchers, and paper writers who want to manage papers, read PDFs, annotate, translate, screen papers quickly, and use AI agents in one seamless workspace. It combines PDF reading, translation tools, note-taking, structured overviews, and AI-driven summaries into a single desktop application, eliminating the need to switch between multiple apps.
Key Features:
Unified Workspace: Combines PDF reading, translation, annotation, and AI-powered summaries into a single interface.
Structured Paper Overviews: Automatically generates summaries for quick screening and later review.
Inline Notes System: Stores rich notes with headings, lists, code blocks, tables, math, highlights, links, folders, and backlinks.
AI-Powered Translation: Pre-translates MinerU-parsed structural blocks for instant access during reading.
Local RAG Knowledge Base: Builds a local knowledge base from imported papers and notes for efficient retrieval.
Zotero Compatibility: Imports Zotero collections, tags, and PDF attachments as an optional feature.
Audience & Benefit:
Ideal for academics, researchers, and paper writers who need to manage large libraries of papers without fragmented workflows. PaperQuay streamlines the research process by enabling users to import, read, translate, annotate, organize, and build knowledge bases in one place, saving time and maintaining focus on their work.
macOS now uses the native traffic-light window controls and hides duplicate custom window buttons.
PDF.js document cleanup is safer when opening, switching, and closing papers, reducing stale worker and render-task errors.
Agent RAG answers can show clickable citation tags that jump back to the referenced paper block or page.
RAG Top-K settings are clearer, and translation popovers now have independent Settings switches.
Side-panel chat input, multi-session QA, and Agent interaction details have been refined.
PaperQuay - Open AI Paper Workspace That Keeps Reading Flow Intact
PaperQuay is more than a PDF reader, AI summary tool, or Zotero add-on. It is a local-first, open-source AI paper workspace designed for graduate students, researchers, and heavy paper-reading users who want to import papers, read PDFs, translate, generate paper overviews, write inline research notes, organize tags, import Zotero libraries, use Agent-assisted literature management, and build a local RAG knowledge base without leaving the same desktop app.
Traditional paper reading often means switching between Zotero, a PDF reader, translation tools, ChatGPT, and a separate note app. PaperQuay brings those steps into one continuous desktop workflow so importing, reading, understanding, translating, annotating, note-taking, organizing, and knowledge-base building can happen in the same place while keeping Zotero compatibility optional rather than mandatory.
Technically, PaperQuay is built as an Electron + React + TypeScript/Vite desktop application. The React renderer implements the literature library, PDF reader, rich notes, Agent workspace, and settings UI; the Electron main process and local Node.js backend handle filesystem access, IPC, Zotero import, SQLite persistence, app updates, and cross-platform packaging. PDF rendering uses PDF.js, rich notes use Tiptap/ProseMirror, local data uses SQLite/sql.js and sqlite-vec, and AI features connect through OpenAI-compatible APIs for translation, paper overviews, Agent tool use, and RAG retrieval.
Research workflow problem
Traditional tools
PaperQuay
Translation latency interrupts reading
Translate only after selecting text, often with visible API delay
Pre-translate MinerU structural blocks and jump instantly to cached translations
Side-by-side translation hurts focus
Two columns require constant eye movement and can break formatting
Keep the original PDF visible while navigating to precise translated blocks on demand
Pure translated files lose source context
Original wording, terminology, and academic expression are hidden
Keep source text, parsed blocks, translation, notes, and overview linked together
Paper notes become detached
Notes live in a separate app and lose PDF position context
Store rich notes, tags, links, paper references, and backlinks in the local library
Fast paper screening is repetitive
Upload PDFs to an LLM one by one and manually organize outputs
Generate and store structured paper overviews inside the local library
AI model choices are locked down
Built-in models or platform-specific token pricing
Bring your own OpenAI-compatible endpoint, model, and runtime parameters
Large libraries are hard to clean
Manual renaming, tagging, metadata fixes, and classification
Agent tools can assist with batch rename, metadata completion, tagging, and classification
Zotero migration is inconvenient
Either stay locked in Zotero or rebuild everything manually
Import Zotero collections, tags, and PDF attachments as an optional source
What Makes PaperQuay Different
Live workflow demo: browse the library, open papers, inspect structured reading, and move into the Agent workspace without leaving the same desktop flow.
Instant Block-Level Translation
PaperQuay uses a translation workflow designed for long paper reading sessions. It can translate and cache MinerU-parsed structural blocks in advance. Later, when reading, clicking a source block can instantly jump to its translated counterpart. Translation no longer needs to happen only after each click or selection.
Tiptap-Based Notes Workspace
PaperQuay includes a dedicated Notes workspace built on Tiptap. Each note is stored locally as Tiptap JSON, rendered HTML, and searchable plain text. The editor supports headings, lists, task lists, code blocks, tables, images, math, highlights, links, slash-style insertions, folders, pin and favorite states, outline, backlinks, and autosave.
Notes are designed to stay inline with the research workflow. You can connect ideas with [[note]] links, organize topics with #tags, reference library papers with @paper, and jump through those inline references instead of keeping reading notes in a separate note silo.
Fast Paper Screening from the Overview Panel
PaperQuay is designed not only for deep reading, but also for screening large numbers of papers quickly. In the overview panel, each paper can directly surface AI-generated fields such as background, research question, method, experiment setup, key findings, conclusions, and limitations.
Reading Time Visibility
PaperQuay records time spent across PDF positions and surfaces it as reading heat previews in the library and a dedicated reading-time chart in the paper detail panel. This makes it easier to see which parts of a paper have actually received attention.
Literature Library, Not Just Import
PaperQuay can build an independent local library with PDF import, a configurable storage folder, categories, tags, metadata editing, search, filtering, notes, and local SQLite persistence. Zotero remains supported as an optional import source, not a required dependency.
Agent Operations for Paper Management
The agent workspace is designed for library operations, not just conversation. It can assist with batch renaming, metadata completion, smart tagging, tag cleanup, automatic classification, and paper summarization while exposing tool calls and results for user review.
Core Workflow
Step
What happens
1. Import PDFs
Drag PDFs into the app or choose files from the import dialog.
[[note]] wiki links, #tag references, @paper references, autocomplete menus, and inline navigation across notes and papers
Reader
PDF reader with MinerU structured block views, region-based linkage, reading heat progress, reading-time recording, and annotation tools
Translation
Full-text translation, cached block translations, and selection translation through OpenAI-compatible models
Paper overview
AI-generated screening fields for background, research questions, methods, experiment setup, findings, conclusions, and limitations
Agent workspace
Conversation UI with execution traces, tool call cards, paper selection, metadata tools, rename tools, tagging, classification, and summaries
Zotero import
Import local Zotero collections, tags, and available PDF attachments from zotero.sqlite
Backup
WebDAV backup and restore for the library database, notes database, and local RAG SQLite database
Updates
In-app update checks, Windows and Linux automatic update flow, and macOS release-page handoff
Themes
Light and dark UI modes optimized for long desktop reading sessions
First-Run Workflow
Open Settings and choose a default paper storage folder.
Import PDFs by drag and drop or by clicking the import button.
Confirm or edit metadata in the import confirmation dialog.
Let PaperQuay copy PDFs into its storage folder and save records in the local library.
Create categories and subcategories from the left sidebar.
Drag papers into categories, add tags, mark favorites, and open papers in the reader.
Open Notes to create rich-text notes, link related ideas, and connect notes to papers.
Configure an OpenAI-compatible endpoint and model if you want AI features.
Configure a MinerU API key if you want MinerU parsing.
Optionally connect a Zotero data directory and import existing Zotero collections and PDFs.
Architecture
PaperQuay uses Electron as its desktop host. The React renderer talks to a local Electron backend through IPC for filesystem access, persistence, Zotero import, PDF handling, and packaging.
Path
Responsibility
src/
React + TypeScript UI, feature modules, state, and frontend services
src/features/literature/
Local literature library UI, import workflow, category tree, and paper details
src/features/reader/
Reader shell, linked reading workspace, settings, and AI reading actions
src/features/pdf/
PDF rendering, overlays, annotation surface, and PDF-specific interactions
src/features/blocks/
MinerU block rendering and structured content views
Zustand state for notes, tags, active note selection, autosave, and workspace errors
src/services/
Frontend bridges to Electron IPC commands
src/platform/electron/
Renderer-side bridge wrappers for commands, events, window controls, and file-drop events
electron/
Electron main process, preload bridge, command backend, packaging helpers, and local persistence
The Notes editor uses the official Tiptap packages and was implemented against the upstream source at ueberdosis/tiptap. The local WikiLink, HashTag, and PaperReference extensions follow the same architecture as Tiptap's official Mention node and @tiptap/suggestion plugin: a Tiptap inline node stores structured attributes, and the Suggestion plugin handles matching, rendering, keyboard navigation, and insertion. The editor also follows Tiptap's React NodeView examples for component blocks: custom blocks are real Tiptap nodes rendered through ReactNodeViewRenderer, with NodeViewWrapper and NodeViewContent separating non-editable controls from editable content.
Requirements
Node.js 18 or newer
Windows, macOS, or Linux
Optional external services:
MinerU API key for cloud PDF structure parsing.
OpenAI-compatible API key for paper overviews, translation, QA, and agent tasks.
Internet access for OpenAlex and Crossref metadata enrichment.
Optional OpenAlex premium API key and mailto polite-pool email for steadier batch metadata lookup.
Development
Install dependencies:
npm install
Start the desktop app in development mode:
npm run dev
Build the frontend only:
npm run build
Preview the built web assets:
npm run preview
Build the desktop installer:
npm run electron:build
Zotero Compatibility
PaperQuay can read a local Zotero data directory that contains zotero.sqlite. During import it copies the Zotero database to a temporary read-only working file and does not modify your original Zotero database.
Imported data enters PaperQuay's own local literature library. Zotero collections become local categories, and available local PDFs inside those collections are copied into the PaperQuay paper storage folder.
Zotero is an optional compatibility source, not a required dependency. You can build a complete library directly inside PaperQuay without using Zotero.
Data and Privacy
PaperQuay is local-first. The literature library, notes, and local RAG indexes are stored in SQLite databases, and imported PDFs are stored in the paper storage folder you configure.
Optional WebDAV backup can upload the local library, notes, and RAG databases to the remote server you configure. API keys, local PDFs, parser outputs, and backups should stay out of source control.
Do not commit local data, API keys, PDFs, parser outputs, notes databases, or backups. The .gitignore excludes common local runtime folders, SQLite databases, legacy JSON library data, API key files, build output, backup archives, and private PDFs by default.
Todo Roadmap
These items are planned or still being deepened beyond the completed features above.
Better metadata extraction from PDF first pages.
DOI / arXiv / Semantic Scholar enrichment options.
Deeper two-way binding between PDF regions, annotations, and standalone notes.
Citation style generation and export.
Folder watching and automatic import queues.
RAG-based knowledge-base QA across papers and notes.
One-click survey generation and Word / LaTeX research draft generation.
Signed macOS release flow for smoother installation and update checks.
Optional cloud sync after the local-first model is stable.
Acknowledgements
PaperQuay is also shaped by discussions, feedback, and shared ideas from the LinuxDo community.
The Notes workspace builds on Tiptap. Thanks to the Tiptap maintainers for the extensible editor framework and examples that help power PaperQuay's note-taking experience.
License
PaperQuay Community Edition is licensed under AGPL-3.0-only.
If you distribute modified versions or provide modified versions over a network, keep the license and copyright notices, mark your changes, and provide the corresponding source code under AGPL terms. For closed-source commercial licensing, commercial support, or brand-name permission, contact the maintainer separately. See TRADEMARKS.md for brand-use notes.
