New · Free · Open Source

Read more.
Forget less.
Think deeper.

A structured literature tracker built for graduate students and researchers. You do the reading. You do the thinking. LitFlow handles the organization — and lets AI synthesize your notes, not replace them.

Single HTML file. No account. No server. Your data never leaves your machine.

Literature review is a slog.
It doesn't have to be.

Most PhD students read sporadically, take notes inconsistently, and spend hours reconstructing what a paper said when it's time to write. AI tools promise to help — but end up doing the reading for you, which defeats the point.

The reading habit that compounds

One paper per day. Six structured note fields. A streak counter that makes consistency feel earned. At 365 papers a year, the habit compounds into a breadth of knowledge that sets your writing and thinking apart.

AI as retrieval, not replacement

LitFlow's synthesis feature works only from your notes — it is explicitly instructed not to add information from its training data. The AI surfaces what you already captured. The thinking remains yours.

Your data, your machine

LitFlow is a single HTML file. It runs in your browser, stores everything locally, and works offline. In Chrome or Edge, you can connect a data file on your hard drive — your library persists even if you clear browser data. No account, no server, no company holding your literature database hostage.

Built by a researcher, for researchers

This came out of a real PhD workflow — MUSC Systems Neuroscience, multimodal behavioral phenotyping, the kind of lit review that spans five fields simultaneously. It solves the problem I actually had.

Everything in one file.

📚

Structured note fields — six, not sixty

Every paper gets the same six fields: Study Design, Key Methods, Main Findings, Relevance to Your Project, Open Questions, and Notes. Consistent structure makes your notes actually useful when you write — and makes AI synthesis dramatically more reliable than unstructured prose.

🗂️

Custom reading tracks

Define your own literature domains — give each a name and a color. Papers can be filtered, sorted, and synthesized by track. Useful whether you're building depth in one field or managing the five-literature problem that most interdisciplinary PhD students face.

LLM synthesis — from your notes only

Select a track and a synthesis mode: draft literature review section, gap analysis, methods summary, key findings table, or a custom prompt. The prompt engineering is built in — every synthesis call instructs the model to work only from your captured notes, not from its training data. Your API key goes directly to your chosen provider; it never touches any server.

Claude · OpenAI · Local LLMs (Ollama, LM Studio)
📎

Quick-add from inside any paper

While reading and taking notes, paste a title, DOI, or URL into the "Additional Readings" field and hit Enter. The paper is instantly added to your library as Unassigned / To Read — no navigation, no interruption. DOIs auto-resolve to title and citation via CrossRef.

🔥

Streak tracking and progress visibility

A day streak counter, per-track progress bars, and a progress line across the top of the page that fills as you complete papers. Small things, but consistency needs feedback loops — especially in the long middle of a PhD where progress is invisible.

📤

Export to JSON and Markdown

Back up your entire library as JSON (re-importable) or export completed notes as formatted Markdown — ready to drop into Obsidian, Notion, a dissertation chapter, or a grant application. The Markdown export is the bridge between your reading habit and your writing.

💾

Disk persistence — survive browser clears

In Chrome or Edge, go to ⚙ Settings → Data → Connect data file. LitFlow links to a .json file on your hard drive and writes every save directly to it. Clear cookies, reinstall the browser, switch computers — your library is still in the file. Firefox and Safari users keep the standard behavior with regular JSON export as backup.

Chrome · Edge 86+

Up and running
in under five minutes.

1

Download the file

Click "Download LitFlow" above. You get a single litflow.html file — no install, no dependencies, no build step.

2

Open it in your browser

Double-click the file or drag it into Chrome, Edge, Safari, or Firefox. All data is saved automatically in your browser's local storage — close and reopen anytime, your notes are still there.

File → Open → litflow.html

⚠️ Protect your library — don't rely on the browser alone.

LitFlow stores your notes in your browser's local storage by default. Clearing "cookies and site data" or resetting your browser will permanently delete everything. Do one or both of the following:

  • Chrome or Edge users: Go to ⚙ Settings → Data → Connect data file to link a .json file on your hard drive. Every save writes to disk — your library survives any browser wipe.
  • All browsers (Firefox, Safari, etc.): Regularly use Export JSON backup in Settings → Data to download a copy of your library. Re-import it anytime.
3

Set up your tracks

Go to ⚙ Settings and define your reading tracks — the literature domains your project spans. Name them, pick a color. This takes two minutes and shapes everything else.

4

Add your first paper and start reading

Click "+ Add Paper," fill in the title and link, assign a track. Then open it, take notes in the six fields, and mark it done. That's the daily loop. One paper. Every day.

5

(Optional) Add an API key for AI synthesis

In Settings, pick a provider — Claude, OpenAI, or a local model running on your machine. Enter your API key. It's stored only in your browser, sent only to the API you chose. Once set, the ✦ Synthesize tab turns your completed notes into draft literature review sections, gap analyses, and more.

AI should make you
smarter, not lazier.

"The goal was never to automate the reading. The goal was to remove every friction point that makes the reading feel impossible on a Tuesday morning before clinic."

— Mike Martino, MD/PhD candidate, MUSC

Every synthesis prompt in LitFlow contains an explicit instruction: work only from the researcher's notes — do not add information from training data. This is a deliberate design choice, not a limitation.

The cognitive work of reading a paper — deciding what matters, connecting it to what you know, noticing what's missing — is where expertise is built. LitFlow keeps that work firmly with you. The AI's job is retrieval and synthesis of what you've already processed, not a shortcut around the processing itself.

🤖 Claude (Anthropic)
💬 GPT-4o (OpenAI)
🏠 Ollama (local)
🏠 LM Studio (local)
+ any OpenAI-compatible endpoint

Mike Martino is an MD/PhD candidate in Systems Neuroscience at the Medical University of South Carolina. His research spans fear sensitization, computational psychiatry, and multimodal behavioral phenotyping — which means he has to maintain a reading habit across five literatures simultaneously.

LitFlow grew out of the FIXATE project — a pipeline for extracting acoustic, kinematic, and facial behavioral features from naturalistic clinical video. The same frustration that drove that pipeline (too many good ideas, too little organized knowledge) drove this tool.

He believes researchers shouldn't have to choose between using AI tools and developing genuine expertise. LitFlow is his attempt to prove you don't have to.

📄 A paper is coming

LitFlow is planned for submission to the Journal of Open Source Software (JOSS) alongside a methods paper on structured AI-assisted literature review in graduate science education. If you use this tool and would be willing to participate in a brief survey study, keep an eye on The Neuro-Humanist for updates.

If LitFlow has been useful to your research, a citation will eventually be possible — and deeply appreciated.

Changelog

v1.9 — July 12, 2026
  • Alphabetical by default — the All Papers list now sorts alphabetically by title (case-insensitive) after any filters/search are applied, on top of the existing track/status/type filters and author/keyword search
  • Edit a paper's title — fix a typo or placeholder title without deleting and re-adding the paper. New Title field at the top of each paper's expand panel, right where every other field (Citation, Link, Notes, etc.) already lives
v1.8.1 — July 8, 2026
  • Fixed a data-loss bug — reconnecting a data file (or reopening LitFlow with one already connected) could silently overwrite newer unsaved local changes with an older version from disk. There was no warning and no way back
  • Conflict detection — reconnecting now compares timestamps, and if your local session has changes the file on disk doesn't, you're prompted to choose which version to keep instead of losing anything automatically
  • Automatic recovery snapshot — before any wholesale overwrite (conflict resolution, JSON import, reset), your prior data is saved to a recovery slot. A "Restore previous session" option now appears in Settings → Data whenever there's something to recover
  • Import confirmation — importing a JSON backup now asks before replacing your existing library instead of doing it immediately
v1.8 — July 2, 2026
  • Configurable synthesis token limits — Settings now lets you set the max output tokens for synthesis, with a per-model reference table (context window, max output ceiling) and Consensus-merge-specific guidance. Fixes a bug where output was silently capped at 1,500 tokens for every model and method, which could truncate longer syntheses mid-sentence
  • Citation export — one-click RIS and BibTeX export for your whole library, plus a per-paper "Cite" button
  • Fixed: the Model field in Settings now correctly remembers a separate value per provider, so switching to a local Ollama model no longer risks sending it a Claude model string (or vice versa)
v1.7 — June 30, 2026
  • Multi-Agent Consensus — run the same synthesis prompt 2–3 times independently, then either merge into a consensus or display responses side by side for comparison
  • Batch + Condense — splits your library into configurable chunks (3–10 papers), summarizes each batch independently, then condenses all summaries into a final synthesis — handles large libraries without hitting token limits in a single call
  • Dynamic token cost estimate shown before running any multi-step job, with Ollama recommended for large libraries to avoid API charges
  • Card expand bug fixed — papers in the All Papers view now expand and collapse correctly (duplicate DOM ID issue when the same paper appeared in both Up Next and the main list)
v1.6 — June 2026
  • Disk persistence via File System Access API — connect a data file in Chrome/Edge that survives clearing browser storage
  • Fixed quick-add to library bug (Enter key and Add button were non-functional)
  • Settings now explains that Claude Pro / ChatGPT Plus don't include API access, with Ollama highlighted as a free local alternative
  • Data safety warning added to Getting Started
v1.5 — Initial public release
  • Six structured note fields, custom reading tracks, streak tracking
  • LLM synthesis (Claude, OpenAI, Ollama, LM Studio)
  • Quick-add by title, DOI, or URL with CrossRef auto-resolution
  • JSON and Markdown export
Free & Open Source

Start reading today.

One file. No account. No server. Works offline. Your notes stay yours.

Download LitFlow GitHub ☕ Buy me a coffee

LitFlow is and will always be free. If it saves you time, a coffee is appreciated and helps fund future tools.