The Zeteyn Manual

Everything you need to go from a blank project to a submission-ready manuscript. No prior experience with research software required.

Account & Sign-in

You sign in to Zeteyn with an account. Two methods are supported:

• Email and password. On the sign-in screen, switch between Log in and Register. To register, enter your email, a password, and a display name.
• Google. Click the Sign in with Google button to authenticate with your Google account. The same button both registers a new account and logs you in.

Once signed in, your projects, Evidence Vault, Skills, and settings are tied to your account.

Five-Minute Quickstart

These five steps will take you from an empty workspace to a draft with a cited source and an AI conversation that you can return to. Everything else in Zeteyn builds on these five actions.

1. 1
   Upload a PDF

   Drag any PDF into the Evidence Vault (the right panel). Zeteyn indexes it in the background. You can keep working while it processes. Once indexed, the AI can search it and you can read it in the built-in PDF viewer.

2. 2
   Save a snippet from the PDF

   Open the PDF in the Reader panel. Highlight a sentence. Drag it onto the Editor in the centre. A small menu appears asking what role this evidence plays: General, Methodology, Arguments, or Key Data. Select one. The snippet is now a hard-linked citation: clicking it in future will scroll the PDF to that exact sentence.

3. 3
   Write something in the Editor

   Click in the Editor (centre panel) and type a sentence or paste existing notes. The editor works like a word processor. Press / to see a list of things you can insert: equations, diagrams, code cells, and more.

4. 4
   Ask the AI a question

   The Chat panel is on the right side of the screen. Make sure the mode at the top says Brainstorm. Type a question about your draft or your uploaded PDF. The AI has read everything in your Evidence Vault and will answer using your own sources, not generic knowledge. To ask about a specific sentence, select it in the Editor first, then type your question.

5. 5
   Save the exchange to your Thought Log

   After the AI replies, look for the [+] button at the bottom of the exchange. Click it. The conversation is saved to your Thought Log: a permanent record of your reasoning that the AI can search in future sessions. Your thinking does not disappear when you close the tab.

Understanding the Layout

Zeteyn uses a four-panel workspace. You can resize or hide any panel. Here is what each one does.

Five AI Modes

The Chat panel has a mode selector at the top. The mode you choose changes how the AI approaches your request, not just what it says, but what tools it uses, what it prioritises, and what it produces. Choose the mode that matches what you are trying to accomplish right now.

Switching modes mid-session

You can switch modes at any point. A common arc: Brainstorm to develop an argument → Edit to write it up → Challenge to find weaknesses → Respond once actual reviewer comments arrive.

Agent in the Editor

Native Codex works with your manuscript through reviewable proposals. It reads the relevant modules, uses enabled skills and tools, and presents each proposed change as a diff in the editor with additions highlighted and deletions marked. You accept, return, or adjust each proposal, and accepted changes remain available in the operation log for rollback.

What the agent can do

• Rewrite or restructure any module. Select a paragraph or a section and ask the agent to improve it. The diff appears inline.
• Reorganize multiple modules at once. Ask the agent to reorder sections, split a dense paragraph, or merge two short ones. Each structural move is logged and reversible.
• Run discipline-panel analysis and insert results. The agent can run analysis in the Lab Log, Chemistry, or other discipline panels and insert the output as a block in your manuscript.
• Edit across the whole document, not just the current module. The agent can work on any module in your project and insert citations where you ask.

One-click full revision

In Edit mode, ask: "Revise this section for clarity" or "Restructure the introduction." Native Codex reviews the manuscript context, uses the relevant skills and tools, and proposes a set of diffs. You can accept all, return all, or review them item by item.

Respond mode: reviewer-reply checklist

Drop a reviewer PDF or paste reviewer comments into Chat. Switch to Respond mode. The agent extracts each distinct comment, maps it to the relevant section of your manuscript, suggests a revision, and drafts a response sentence. A checklist appears. You work through it, resolving each item. When all items are resolved, the response letter is assembled and ready to send.

Talking to the AI

The AI understands plain language. You do not need special syntax. Just describe what you want. A few things that help:

• Select text first, then ask. If you want the AI to help with a specific sentence or paragraph, select it in the Editor before typing your message. The AI will know you are referring to that selection.
• Be specific about what kind of help you want. "Improve this" is less useful than "make this argument more concise without removing the claim about sample size."
• The AI will ask for clarification if needed. If your request could mean more than one thing, the AI will describe its interpretation and ask you to confirm before acting.
• The AI never changes your manuscript without showing you first. Any proposed edit appears as a preview. You decide whether to accept it.

Floor buttons: [+] and [+↑]

Each AI exchange is called a "floor." At the bottom of each floor you will see:

• [+]: Save this specific exchange to your Thought Log
• [+↑]: Save all exchanges from the beginning of the session up to and including this one
• The checkbox on the left of each floor lets you select multiple floors; the [+] button then saves all selected floors at once

What the AI can see

Each session provides the selected workflow with the context it needs:

• Your Research Anchor: your core research question, confirmed findings, and constraints (from RESEARCH_ANCHOR.md)
• The current conversation history (recent exchanges in full; older ones as summaries)
• Brainstorm retrieves real records from your Evidence Vault and connected academic indexes before synthesis
• Native Codex Edit reads the relevant manuscript modules through the White-Box tools before preparing proposals

Connected academic indexes include OpenAlex, Crossref, PubMed, Semantic Scholar, and arXiv. Retrieved records become the constrained evidence for Brainstorm synthesis.

Thought Log

The Thought Log is a permanent record of your reasoning. Unlike your conversation history (which compresses over time), the Thought Log is persistent across sessions and fully searchable.

To view your Thought Log, click the log file in the Left Sidebar or click the [↗] icon in the conversation header. It opens in the Viewer panel alongside your Editor.

What belongs in the Thought Log

Save exchanges where you worked something out: a decision about methodology, a finding you confirmed, a direction you ruled out. The AI can search your Thought Log in future sessions, so saving "why you decided against approach X" is as useful as saving "what you concluded about approach Y."

Research Anchor

The Research Anchor (RESEARCH_ANCHOR.md in your project root) is a special file that the AI reads at the start of every session. It keeps the AI oriented around your core research question even as conversation histories grow and compress.

You can edit it directly or ask the AI to update it: "Add our finding about sample size to the Research Anchor." It contains:

• Your core research question
• Confirmed findings (with their source references)
• Methodological constraints
• Disproven hypotheses (the "Context Blacklist": the AI won't repeat these mistakes)

Checkpoint & Rollback

Zeteyn automatically saves a checkpoint before every significant edit: a manuscript change, a Thought Log insertion, a bulk replacement. If you accept an AI suggestion and then change your mind, you can roll back to exactly the state before that edit.

Where to find checkpoints

Look at the right edge of the screen. You will see a vertical strip of small round dots, one per conversation floor. Floors that have a checkpoint attached are shown in a different colour (indigo), the same round shape as the other dots, not a special icon. The current floor is a solid white dot. Hover over any dot to see the floor number, and for a checkpoint the reason it was created and when.

Rolling back

Collaboration & Workspaces

Zeteyn supports real-time, multi-user editing: several people can work in the same manuscript at once and see each other's changes live. Shared projects live in workspaces, and each member has a role that controls what they can do.

GitHub Sync

You can connect your own GitHub account and sync bidirectionally: pull a repository into your project, edit it in Zeteyn, and have the agent commit changes back, all under your own GitHub identity, to your own repositories.

Evidence Vault

The Evidence Vault is your indexed library of sources. Everything you add to it is searchable by the AI and linkable from your manuscript. Think of it as the verified foundation your writing stands on.

Adding sources

• Drag a PDF into the Vault panel. Zeteyn indexes it automatically.
• Paste a URL into the Editor or Chat. Zeteyn captures and indexes it.
• Paste BibTeX anywhere. Zeteyn detects it and creates a citation entry automatically.
• Connect Zotero (see Zotero Integration) to sync your existing library.
• Ask the AI: "Search ArXiv for recent papers on [topic]." Any paper you save from the results goes into the Vault.

Source Cards

When Brainstorm cites a retrieved paper, the matching Source Card appears beside the relevant discussion with the paper title, author, year, and verification state. Unmatched citations receive a visible verification marker. Two actions appear on each matched card:

• Save to Vault: downloads and indexes the paper if not already in your Vault
• Open in Reader: opens the paper in the Viewer panel alongside your Editor

Snippets

A snippet is a piece of text from a source (a sentence, a table, a passage) that you have saved and linked to its origin. Every snippet remembers exactly which PDF it came from and which page.

To create a snippet: open a PDF in the Reader, highlight text, and drag it onto the Editor. You will be asked to categorise it (General, Methodology, Arguments, or Key Data). The snippet appears in your manuscript as a citation that links back to the source.

Literature Search

Zeteyn searches external academic databases in addition to your own uploaded sources. You can ask the AI to search, or use the literature search directly. Covered sources include:

• arXiv: preprints across physics, maths, CS, and more
• PubMed: biomedical and life-sciences literature
• Semantic Scholar: cross-disciplinary index with citation graphs
• OpenAlex: open catalogue of scholarly works
• Crossref: DOI registry and metadata
• IEEE: engineering and computer-science publications

Where a paper has an open-access version, Zeteyn resolves the full-text PDF so you can read it in the built-in reader. Save any result to your Evidence Vault to index it and cite from it.

PDF Reader

The Reader panel opens any PDF from your Vault alongside your Editor. The two panels are connected: a snippet you drag from Reader to Editor creates a permanent link: click the citation in your manuscript and the Reader scrolls to that exact sentence.

To verify any citation later: click the citation marker in your manuscript. The Reader opens the source PDF at the exact sentence. No more hunting through reference lists.

The Reader also supports annotation: you can highlight, underline, add comment notes, or place free-text notes on a PDF, and export a highlighted passage straight to your Evidence Vault as a snippet.

Built-in Browser & Web Clipper

Zeteyn includes a built-in browser so you can look things up without leaving the app. Open it from the Activity Rail; it opens on a research-friendly start page.

The Web Clipper lets you capture what you find on a page and bring it into your project, where it can be turned into evidence and cited like any other source.

Snippets & Citations

Snippets are the granular unit of evidence in Zeteyn. They live in your Evidence Vault, each one linked to a specific location in a source document.

Inspiration snippets

Not all useful fragments come from sources you have uploaded. If you encounter a thought you want to keep but cannot yet attribute, use Save as Inspiration (available in the Chat right-click menu and the Editor right-click menu). Inspiration snippets are saved without a source link: a note to yourself rather than a citation. You can add a source link later.

Automatic detection

When you paste certain formats into the Editor, Zeteyn recognises them and offers a matching action:

• URL → offers to import it as a source
• BibTeX block → offers to insert a citation and continue in the citation workflow
• CSV / table-like data → offers to create a structured data artifact through chat
• SMILES notation (e.g. C1=CC=CC=C1) → offers to create a molecule artifact through chat

Zotero Integration

If you already have a Zotero library, you do not need to re-upload anything. Connect Zotero once and your full library syncs into the Evidence Vault, including your annotations.

Once connected, your Zotero items are available in the Evidence Vault. (Mendeley is not currently supported.)

The Editor

The Editor is where your manuscript lives. It works like a familiar word processor, with a few additions specific to research writing.

Modules

Your manuscript is divided into modules: sections like Introduction, Methods, Discussion. Each module is a separate file in the manuscript/ folder of your project. The Left Sidebar lists them in order. You can reorder, split, or merge modules from the sidebar.

The status bar

At the bottom of the Editor, a floating status bar shows your word count (total, current module, and selected text), the current language, and the connection status. It disappears when you start typing and reappears on hover.

Accepting AI suggestions

When the AI proposes an edit in Edit mode, a diff preview appears: additions highlighted, deletions marked. You can:

• Click Accept to apply all changes
• Click Discard to reject all changes
• Click individual changes in the diff to accept or reject them one by one

After accepting, the Editor scrolls to the first changed location, highlighted briefly. If you do not like the result, the Undo button on the toast notification rolls back to the checkpoint created just before the edit.

Command Palette

Type / in the Editor to open the slash-command menu. Keep typing to filter the list. These commands insert content blocks or start an AI action at the cursor.

Right-Click Menu

Select any text in the Editor or the Reader panel and right-click to see context-sensitive actions. These are the same actions you can request verbally from the AI: the menu is a vocabulary list, not the only way to trigger them.

Keyboard Shortcuts

Challenge Mode

Challenge mode reads your manuscript as a hostile peer reviewer would. It audits every claim, cross-checks your citations against what the cited papers actually say, constructs counterarguments, and produces a structured issue report. Use it before you submit, or before someone else does this to you.

Challenge mode also works on external papers. Load a paper into the Viewer panel, switch to Challenge, and ask Zeteyn to audit it. Useful for dissertation literature reviews, grant committee preparation, or systematic critique.

Respond Mode

When you receive reviewer or supervisor comments, Respond mode turns them into a structured checklist of manuscript changes and drafts your response letter alongside each revision.

Peer Review Mode

Peer Review mode produces a formal structured review of an external paper. It follows a four-step protocol, uses standard academic sections, and archives the report for later access and HTML export.

The four-step protocol Peer Review runs: (1) full structural scan of the paper's argument arc, (2) statistical and methodological audit, (3) citation verification against known literature, (4) structured verdict with actionable recommendations.

Pre-Submission Check

Before you submit, run the pre-submission check to catch problems while you can still fix them. The results appear as a fix-list. Work through each item and mark it resolved. Items you choose to leave unfixed remain visible so you can make an informed decision before submitting.

What it checks

The pre-submission check runs a rule library against your manuscript and flags each rule as compliant, needs-review, or non-compliant. The rules are grouped as follows:

These checks work by scanning your modules for the required elements, so treat a "needs review" result as a prompt to look, not a definitive pass or fail.

How to run it

Skills

A Skill packages a reusable system prompt (a set of instructions plus an allowed tool list) that you can reapply instead of re-explaining the same workflow every time. Skills are stored in your account (not as files on disk) and managed in the My Skills panel.

Creating a Skill from a conversation

The fastest way to make a Skill is to save one from a chat you already like. Open the conversation menu and choose Save as Skill. Zeteyn summarises the conversation into a draft: a name, a short and long description, a system prompt, and tags. Review and edit those fields, then save. The Skill is created as a private draft by default.

Managing and sharing Skills

In the My Skills panel you can edit, delete, or change the visibility of each Skill. You can also open the Skill Marketplace to browse Skills shared by others, and publish your own. Zeteyn does not ship pre-installed Skills: the library is what you and the community create.

Discipline Panels

Beyond the core writing tools, Zeteyn ships discipline-specific panels. Open them from the Activity Rail: the icon strip on the right edge of the screen; less-used panels sit under the "..." expander, and some are launched as workspace tools.

Data & Code Cells

Zeteyn has live code cells that run in your browser: no server round-trip and no local install. Python runs via Pyodide and R runs via WebR (both compiled to WebAssembly). Cells work like a notebook cell embedded in your work.

You can ask the AI to write or debug code for you: describe what you need in Chat, or select a cell and switch to Edit mode.

Molecules, Maps & Diagrams

Molecule structures

Zeteyn can fetch and display a molecular/protein structure summary from the Chemistry and Data Science tools: identifiers, chain and atom counts, organism, resolution, and a link out to the source database (RCSB). Pasting a SMILES string in the Editor offers to create a molecule artifact through chat. Note: a fully interactive rotate-and-zoom 3D viewer is not embedded in the app today; the molecule block is a structure summary, not a live 3D scene.

TikZ Diagrams

Type /tikz to insert a TikZ diagram block, then write your TikZ code. The output is a vector graphic that scales cleanly for any page size.

Geospatial Maps

Geospatial maps live in the Spacetime panel (Leaflet-based), which you open from the Activity Rail, not from a slash command. Use it for studies with a geographic dimension.

Equations

Insert a LaTeX block with the /latex slash command to write equations. LaTeX renders to a typeset equation.

Remote Lab Connection

If your simulations or data processing run on a lab server or HPC cluster, you can connect over SSH and pull results into your project.

The SSH panel is a form-and-file-transfer tool, not an interactive terminal. For batch computing it can poll HPC job status on SLURM, PBS, and SGE schedulers, so you can check on a submitted job and pull its outputs once it finishes. Pulled files land in your project, where you can open them in the Viewer or reference them in your manuscript.

Showroom & Formatting

The Showroom is a preview of your manuscript in a specific journal's layout. It is not the editor: it is a read-only preview that shows you exactly how your manuscript will look when exported.

Switching templates does not change your manuscript content, only the presentation. You can switch between templates as many times as you like and export the same manuscript in multiple formats.

Export Options

Export your manuscript through the Export wizard. Six output formats are available:

Desktop App: Download & Install

Zeteyn desktop is available for Windows, macOS, and Linux. The desktop app runs a local AI model on your own hardware and keeps your files on your own disk, which is the main reason to choose it over the web app. All three installers are reachable from the download buttons on the homepage, or directly at zeteyn.com/download.

Windows

Download Zeteyn-Setup-0.1.6-win-x64.exe and run it. The installer is unsigned, so Microsoft Defender SmartScreen shows a "Windows protected your PC" message on first run. Click More info, then click Run anyway to continue with setup.

macOS

Download Zeteyn-0.1.6-arm64.dmg. This build targets Apple Silicon (M1 and later). The app is unsigned, so Gatekeeper asks you to confirm before it opens for the first time.

Linux

Download Zeteyn-0.1.6.AppImage.

Your AI Provider

Zeteyn runs on a single engine (Codex) with three interchangeable "fuels" you choose in Settings → Tools. Under AI Fuel you pick one of three cards: A · My API key (bring your own provider key), B · My local Codex (route to a Codex session running on a computer you own), or C · Ready-to-go (the hosted Zeteyn Pro sandbox, no setup). The card you select powers your chats.

Option A: Add a provider key

Go to Settings → Tools, open the AI Fuel section, and use the A · My API key form. Select your provider, paste your API key, and click Save API Key. There is no separate "Test Connection" button: saving stores the key, and the key is used the next time you chat.

Option B: Use your local Codex

Fuel B routes your chats through the Codex CLI running on a computer you own (a "home runner"). The sign-in to your ChatGPT/OpenAI account happens locally in your own terminal: the web app cannot perform that OAuth, and never sees or stores your Codex session. Setup is a few terminal commands on your machine, then a one-click detection in the app.

Option C: Ready-to-go (Zeteyn Pro)

The C · Ready-to-go card is the hosted Zeteyn Pro sandbox: it uses platform model credentials, so there is nothing to install or sign in to. Upgrading to Pro is done from the same card.

Privacy & Offline Search

Zeteyn's semantic search runs locally by default. It uses a small bundled embedding model (BAAI bge-small) that runs on your CPU, downloaded once to a local cache: no cloud embedding service and no API key are required to index and search your sources. This means the content of your Evidence Vault does not have to be sent to a third-party embedding provider just to make it searchable.

You keep the choice of AI provider (see above): a hosted plan, your own API key, or your own local Codex runner. BYOK and local-runner setups keep the model calls under your own account.