Read papers in your language. Let AI manage your library.
中文 · Download · Development
Zoro is an AI-native literature management tool built for non-English-speaking researchers and the age of AI agents. From the very first line of code, native-language reading and AI agent collaboration are core design principles, not afterthoughts.
Cross-platform desktop app (macOS / Windows / Linux). Local-first. Your data stays yours.
Download the installer for your platform from the Releases page:
| Platform | File |
|---|---|
| macOS (Apple Silicon) | Zoro_x.x.x_aarch64.dmg |
| macOS (Intel) | Zoro_x.x.x_x64.dmg |
| Windows | Zoro_x.x.x_x64-setup.exe |
| Linux (Debian/Ubuntu) | Zoro_x.x.x_amd64.deb |
| Linux (AppImage) | Zoro_x.x.x_amd64.AppImage |
Want to build from source? See the Development Guide.
Most reference managers treat translation as a checkbox feature. Zoro weaves native-language reading into every surface of the app.
Three display modes, one click — Original / Bilingual / Translated. Applies globally across paper lists, abstracts, and detail views. Titles and abstracts are auto-translated with an immersive bilingual layout — translation displayed prominently, original below.
Side-by-side bilingual PDF reader — Original PDF on the left, translated PDF on the right, synchronized scrolling. Both sides support highlights, annotations, and ink notes.
Full-text HTML translation — ArXiv HTML papers are translated paragraph by paragraph in the background, with real-time progress. Read while it translates.
Bilingual subscription feeds — Browse HuggingFace Daily Papers and other feeds in bilingual mode to quickly triage what's worth reading.
All translations are cached locally — open a paper again and it's instant.
Zoro is designed from the storage layer to the protocol layer for AI agents.
Built-in AI Assistant — In-app agent panel with multi-turn conversations. Ask questions, summarize, translate, or analyze the current paper directly. Supports image input, tool calls, and chain-of-thought display.
MCP Server — Built-in Model Context Protocol server with ~35 tools. Claude Desktop, Cursor, OpenCode, and other AI tools can search, browse, and manage your papers. One toggle in Settings to enable. See the MCP Server docs for details.
The Filesystem Is the API — Every paper lives in its own directory with a structured metadata.json. AI agents can read your library by simply navigating the filesystem — no special SDK needed. The attachments/ directory is agent-writable, allowing agents to generate summaries, translations, and reports.
~/.zoro/library/papers/
2017-attention-is-all-you-need-a1b2c3d4/
metadata.json ← Structured metadata, directly readable by agents
paper.pdf ← PDF full text
abs.html ← HTML full text
attachments/ ← Agent-writable
summary.md ← AI-generated summary
translation-zh.md ← AI-generated translation
notes/ ← User notes
- Zotero Import — Import your existing Zotero library including papers, collections, tags, and metadata. Migrate seamlessly without losing any data
- WebDAV Sync — Sync your library across devices via any WebDAV service. Conflict-free, encrypted, you own the server
- Full-featured PDF Reader — Highlights, underlines, sticky notes, freehand ink, outline navigation, citation hover preview
- Browser Extension — Chrome extension to save papers with one click from ArXiv, DOI pages, and any academic site. Zotero Connector compatible
- Import & Export — BibTeX / RIS format support, drag-and-drop PDF import, formatted citation output (APA, IEEE, MLA, Chicago)
- Local-first & Private — SQLite database, all data stored locally, works offline. Paper data never leaves your machine
- Cross-platform — Native on macOS, Windows, and Linux