PaperZorro

What problem am I addressing?

Keeping up with the literature is a grind. New papers land on PubMed, arXiv and bioRxiv/medRxiv every single day, and the usual routine — running the same searches over and over, skimming dozens of abstracts, and manually adding the few relevant ones to Zotero — eats time you’d rather spend actually reading. Worse, it’s easy to miss things: a search you forgot to re-run, a preprint that slipped through.

PaperZorro is built to take that grind off your plate. It’s an AI-powered literature scout that lives inside Zotero. You compose a precise boolean query, pick your sources, and a language model triages every candidate paper against your search intent — labelling each one YES or NO with a short reason. You review the verdicts, check the papers you want, and import them as proper Zotero items in one click. Save the query as a schedule and PaperZorro keeps scouting on its own, so newly published relevant work shows up in your library without you lifting a finger.

By default the AI filtering runs locally via Ollama — no API keys, no paper data leaving your machine. Cloud providers (OpenAI, Anthropic/Claude, Gemini) are available in Beta if you prefer.

How it works

PaperZorro adds a panel to Zotero with four tabs — Settings, Search, Results and Schedule — that walk you through the full scouting workflow:

Settings — choose your AI provider

Pick an LLM provider and model, and tune the relevance filter prompt to match exactly what you care about. By default PaperZorro talks to a local Ollama instance — keeping AI filtering local means no API keys and no paper data ever leaves your machine.

Search — build your query

Use the Builder to assemble nested AND / OR / NOT groups with a point-and-click interface, or switch to Raw for a hand-written query string when you need full control. Queries are automatically translated to each source’s dialect. Choose a source (PubMed, arXiv or bioRxiv/medRxiv), set a date range and result cap, and watch the live progress bar as PaperZorro fetches and AI-filters each paper.

Results — review and import

Browse the AI verdicts. Each paper is tagged YES or NO with a short justification, and you can expand the abstract to read more. Pick a target collection (any collection or sub-collection in your library), check the papers you want, and click Import Checked to Zotero — they become proper Zotero items with full metadata.

Schedule — automate discovery

Save a query as a recurring schedule. Give it a name, set the source, interval (e.g. weekly on Monday at 08:00) and a max-paper cap. PaperZorro then runs the search on schedule so newly published relevant work shows up in your library without you lifting a finger.

Key Features

• Visual query builder — Compose nested AND / OR / NOT groups point-and-click, or drop into Raw mode for full control
• Multi-source search — PubMed, arXiv and bioRxiv/medRxiv, with per-source syntax translation
• AI relevance filtering — Each candidate paper is judged YES/NO against your search intent with a short reason
• Local-first — Runs against Ollama on your own machine by default — no API keys, no data leaving your computer
• Cloud providers (Beta) — Optional OpenAI, Anthropic/Claude, Google Gemini, and any OpenAI-compatible endpoint
• Custom filter prompt — Tune the relevance instructions to your lab’s exact criteria
• One-click import — Approved papers become proper Zotero items with full metadata in the collection of your choice
• Scheduled searches — Save a query to run daily or weekly so your library stays current automatically

Requirements

• Zotero 7.0 or later
• (Optional) Ollama for local AI filtering — cloud providers can be used instead