From the team that built Overleaf·Closed beta · June 2026

LaTeX + python

Tartu is an AI-first LaTeX environment for the documents that earn grants, get cited, and ship to peer review. Compile in milliseconds. Cite from your own library. Run pandas next to the paragraph it informs. Write together, without losing a keystroke.

Six things, done right.

Everything in Tartu is built around one rule: the document is the source of truth. The agent, the compiler, the data tools, the history — they all serve the page, not the other way around.

01Editor

A LaTeX editor that respects the document.

aims.tex
refs.bib
methods.tex
14\section{Specific Aims}
15
16Chronic kidney disease (CKD) affects [1] over 37
17million Americans, with marked disparities in
18outcomes across rural cohorts [2].
19
20\subsection{Aim 1: Identify risk}
21
compiled · 142ms
02Compile

Compile in milliseconds. Not minutes.

$ pdflatex aims.texlive · running
Tartu incremental0ms
parsing tree · resolving refs…
pdflatex classic0.0s
still compiling… (typesetting page 3 of 47)
median over last 100 builds59× faster
03Tatu

An agent that reads your sources, draws your diagrams, fixes your build.

Chat · groundedAI · 4 sources
You

Sketch the study design as a CONSORT diagram and pull the cohort numbers from Hwang 2024.

04Notebook

Python on your pilot data, embedded in the manuscript.

eGFR_decline.py · cell 3 / 7scipy · pandas · matplotlib
# Pilot cohort — 36-month eGFR trajectory
import pandas as pd; from scipy import stats
df = pd.read_csv("pilot/ckd_egfr.csv")
fit = stats.linregress(df.month, df.egfr)
eGFR decline · n=412
observedlinear fit
Lookback 36mo
Cohort CKD 3b+
Fit
β = −2.31 mL/min/y
R² = 0.74 · p < 0.001
05Collaboration

Co-write with your team. Unlimited seats, no friction.

aims.tex · 3 editingMPDCY
% Specific Aims · NIH R01 · revision 12
\section{Specific Aims}
Chronic kidney disease (CKD) affects over 37 million
Americans, with marked disparities in outcomes across
rural cohorts. Our central hypothesis is that
\textbf{site-tailored monitoring} reduces eGFR
decline by 18% at 36 months.
\subsection{Aim 1} Identify modifiable risk
factors in a 6-site longitudinal cohort.
Maya Patel
Daniel Cho
You
06History

Every keystroke kept. Forever.

git · mainretention · ∞

Unlimited version history — included.
9f3a2eAim 3: tighten power calc2 min
7d1b0cHwang 2024 → ref 12; fix bib26 min
4e90afTatu drafted §Significance1h
2a87deMaya: pilot data table 34h
1c50d2initial commit · grant skeletonMar 2
auto-snapshot on saverestore any revision · diff any range

A first for LaTeX

matplotlib, plotly & Python
inside your LaTeX project.

No more exporting a PNG, re-uploading, recompiling. Write the code that makes the figure in the same project as the paper — run it, and the plot lands in your document by reference. Change the data, the figure updates. This is the part nobody else has.

matplotlibfigures.py
by site

The plots you already know. Publication-grade figures, rendered straight into the manuscript.

\includegraphics{fig:enrollment}
plotlytrajectory.py
treatmentcontrol

Interactive, multi-series charts. Explore live, then freeze the exact view into a figure.

\includegraphics{fig:egfr-arms}
pythonanalysis.py
observedlinear fit

pandas, scipy, statsmodels — the full stack runs in a sandbox next to the paragraph it informs.

\includegraphics{fig:fit}
run in a sandbox · pandas · scipy · matplotlib · plotlyembed by reference · recompiles on change

Real-time collaboration

Write together, on one document.

Three people, three cursors, one source of truth. Tartu's Loro CRDT keeps every keystroke in sync — presence, selections, and live cursors, with unlimited seats and no lost edits. Co-write a grant the way you'd sketch on a whiteboard, only it compiles.

Loro CRDT · presence & selectionsunlimited collaborators · zero merge conflicts

From an interview · R01 writer

I spend hours manually checking the LLM's citations. None of these tools give me a direct link from the claim to the page.

— Medical researcher, R01 grant cycle, Feb 2026

By the numbers

142ms
Median compile
59×
Faster than pdflatex
100%
Citations verified to page
History retention

Pricing · One product, three editions

Pay for the agent, not the editor.

The editor, Turbo pdfTeX compiles, and collaboration are unlimited at every tier — even Free. What you pay for is the agent: how many runs you get, the models behind them, plus Python hours, the compatibility compilers (pdfLaTeX, XeTeX, LuaTeX), history, upload room, and your choice of EU or USA workbook residency.

01Open

Free

$0forever

For coursework, exploration, and second readers. No payment, no expiry.

  • +Unlimited collaborators
  • +48 hours of history
  • +Turbo pdfTeX — unlimited compiles
  • +30 mins of Python / month
  • +20 MB file upload limit
  • +50 AI agent runs / month
  • +Community support
Start free

No credit card required

03Professional

Professional

$50/ month

For PIs and grant-writing labs. The full agent, every keystroke, hosted where you need it.

  • Everything in Researcher
  • +Unlimited AI agent runs
  • +State-of-the-art models
  • +24 hours of Python / month
  • +10 GB file uploads
  • +EU / USA workbook residency — your choice, per project
  • +Dedicated onboarding & SLA
Choose Professional

Lab plans available · contact for 5+ seats

50% off student discount
Student discountResearcher plan available for $20/month
One product · three editions · same editorFirst month free · Cancel anytime

Waitlist · Closed beta

The first 50 R01 writers get access at launch.

Closed beta opens late June 2026. Monthly notes from building Tartu — no spam, unsubscribe with one click.

By joining you agree to receive monthly emails. We don't share your address. Unsubscribe anytime. Privacy →

SEATS · 47 / 50 OPEN·3 invited this week

TARTU · VOL. I · NO. 1

USETARTU.COM · PRIVACY · X · CONTACT