Nodus + Zotero | Open Source Research Workspace

01 · The relationship

Nodus does not replace Zotero.

A reference manager and a research workspace answer different questions. Zotero answers what do I have, and how do I cite it. It collects items from the web, holds attachments, syncs across your machines, manages groups and drives citations in your word processor. Nodus does not do any of that better, and it does not try to.

What Nodus answers is what does this material say, and how do the pieces relate to each other. It reads the works in your library, records ideas with the passage and page that support them, connects those ideas with relations that mean something, and gives you a corpus you can search semantically, argue with and write from.

In practice this means you keep your Zotero workflow exactly as it is. Keep saving with the Zotero Connector, keep organising in collections, keep syncing to zotero.org. Nodus reads what is already there, and because it is open source and local-first, it reads it on your own machine.

The connection is read-only, by design. Nodus talks to Zotero through the local API that the Zotero desktop application exposes, and never writes to it. It does not modify your items, does not create collections in Zotero, and never opens zotero.sqlite directly. Attachments are copied into the Nodus Library, and the files in your Zotero storage are only ever read.

02 · Connecting a library

Mirroring your research library into Nodus.

Nodus keeps one cross-vault Library, and a Zotero library can be mirrored into it with its structure intact. Both applications run on the same computer, and nothing is routed through an external service.

01

Have Zotero open

Nodus reads from the local API served by the Zotero desktop application, which means Zotero 7 or newer running on the same machine. There is no API key to paste and no round trip through zotero.org, because the connection is a local one between two applications on your own computer.

02

Choose the library and what to bring

Your personal library and any group libraries you belong to are both available, and Nodus previews a library before importing it so you can see what a sync would bring in. Collections keep their hierarchy, and items keep their stable Zotero keys, which is what lets a later sync recognise the same item rather than duplicating it.

03

Metadata, attachments and notes

An import carries bibliographic metadata, attachments and the child notes attached to your items. Nodus preserves each original attachment and builds a clean reading copy beside it, along with extracted figures, tables and page mappings, so a claim can always be checked against the page it came from.

04

Keep it up to date

Later syncs are incremental: Nodus asks Zotero what has changed since the version it last saw, and picks up new and edited items along with the ones you deleted, rather than re-reading the whole library each time.

05

Decide what belongs to which project

A mirrored Zotero library sits in the shared Library, not inside one project. Linking an item into a vault is what places it in that project's corpus, so the same paper can support a thesis and a course without being stored twice. Unlinking it leaves the global source untouched.

Zotero is not the only route in. The same Library accepts RIS, BibTeX and CSL JSON exported from Mendeley or another manager, plain local documents, and pages captured by the Nodus Connector for Chrome. If you do not use Zotero, nothing on the research page becomes unavailable to you.

03 · The workspace around your sources

What happens to a Zotero library once it is in Nodus.

This is the part Zotero was never meant to do, and the reason the two tools sit well together.

Ideas with their evidence

Works are analysed into five kinds of unit, a claim, a finding, a construct, a method or a framework. Each one keeps the excerpt that supports it, plus the page when the source provides one.

How ideas and evidence work

A graph of your reading

Ideas are joined by relations that mean something, such as supports, contradicts, refines, extends or applies to. They gather into themes you can browse as an interactive research map.

Semantic search over the library

Exact and semantic retrieval across works, ideas, passages and authors, including across languages, with the embeddings computed by a provider you choose or entirely on your own machine.

Debates, gaps and coverage

Where your sources disagree, which questions they leave open, and how well the collection you have actually covers the themes your work depends on.

Deep Research reports

A queued job that pulls the scoped corpus into a long-form report whose citations open the source behind them. Treat it as a draft you can audit rather than a finished text.

Citations in your own style

The citation manager uses real CSL styles and will format with custom .csl files copied over from Zotero, locally on your machine.

Writing beside the sources

Projects, drafts and source-linked notes live in the same workspace, with companions for Microsoft Word and LibreOffice for the manuscript itself.

Translating attachments

Nodus Translate works with text, files and Zotero attachments using the model you choose, preserving DOCX and EPUB structure and offering a facsimile PDF mode that keeps the original layout.

04 · The plugin

Nodus for Zotero, inside Zotero.

Some work never leaves the reference manager. For that, Nodus ships a separate Zotero plugin. It is a standalone add-on that runs entirely inside Zotero and does not need the Nodus desktop application at all.

The plugin indexes the PDF, EPUB and HTML attachments in your library and adds a research sidebar that answers questions across them. Answers come back with the exact passages they rest on and the pages those passages sit on, so you can check a reply instead of having to take it on trust.

Retrieval combines keyword and semantic search and works across languages. The embeddings can be produced by a managed local model that runs on your own machine inside the Zotero profile, or by a provider you configure yourself. Anthropic, OpenAI, OpenRouter, Groq, Cerebras, DeepSeek and Google Gemini are all supported, among others, alongside local Ollama and LM Studio endpoints.

Reading what a scan does not say in text

Vision-capable models can read document pages as images, transcribing scanned text and describing figures, tables and formulas so that material a plain text layer misses still enters the index.

Auditing your own claims

An evidence audit takes claims and checks them against the indexed library, highlighting the ones whose support looks thin. Read it as a nudge to go back to the sources, not as a verdict on whether a claim is true.

Installing it

Download nodus-zotero.xpi from the latest Nodus release. In Zotero, open Tools → Add-ons, choose Install Add-on From File and select the downloaded file. The index it builds stays in your Zotero profile.

05 · Choosing

Which one do you actually need?

The three pieces do different jobs, and most people end up using more than one.

You want to…	Use	Why
Collect references, store PDFs, sync and cite in a word processor	Zotero	That is what a reference manager is for, and Nodus is built on the assumption that you keep doing it there.
Search your library's full text and get answers with page citations, without leaving Zotero	The Nodus plugin for Zotero	It runs inside Zotero, indexes your attachments in your own profile and needs no separate application.
Turn a body of literature into ideas, relations, debates, gaps and a cited draft	The Nodus desktop app	That analysis needs a corpus with its own structure, which is what the academic vault is.
Do all of it	All three	Collect in Zotero, mirror into Nodus for the analysis and writing, and keep the plugin for quick retrieval while you are reading.

Keep going

Nodus and Zotero