you have data and a question. you can't code. your university says avoid us clouds.
That's exactly who MakeMode Notebook is for. Drop a dataset, ask your question in plain language, and get a real, reproducible analysis — computed on European rails, with a strict mode where nothing ever leaves your browser at all.
how it works.
No terminal, no environment to set up, no account required to start.
drop a csv
Any spreadsheet export — survey results, lab measurements, a study log. It's read on your own device; nothing uploads while you're looking at it.
get a working tool
Ask your question the way you'd ask a colleague — "how does satisfaction differ by age group?" — and watch real Python (pandas, NumPy, Matplotlib) write and run the analysis in front of you.
keep or share it
Export the whole session as a real, standard .ipynb file today — yours to keep, no lock-in.
pick your data's security level before you drop a single row.
Labelled to match TU Delft's own data classification, so the tier you pick means the same thing your data steward already uses.
Human-subjects or personal data — anything that identifies an individual.
data never transmitted — check the network tabConfidential or competition-sensitive institutional data.
eu compute, never stored — designed, shipping nextPublic or otherwise non-sensitive data.
eu-stored & shareable — designed, shipping nextToday's working prototype delivers the Critical tier in full: your dataset is read and analysed entirely inside the browser tab (WebAssembly) and never transmitted. Don't take our word for it — open your browser's Network panel while you use it, or read the full safety brief written for the person who has to sign off.
makemode notebook vs claude science vs google colab.
Anthropic launched Claude Science on 2026-06-30 — a real, well-reviewed research workbench, and worth taking seriously. Google Colab has been the default free notebook for over a decade. Here's an honest, sourced look at where each stands today, including where the others are genuinely better.
| makemode notebook | claude science | google colab | |
|---|---|---|---|
| install | None — any modern browser, including Windows | Beta app, macOS 13+ or Linux only, ~5 GB; no Windows at launch | None — browser-based, any OS |
| where your data goes | At the Critical tier: nowhere — computed in-browser, verify it in the Network tab | Your prompt and the model's response are sent to Anthropic's US servers for every call | Runs on a Google Cloud VM; the region isn't guaranteed, and AI-feature requests aren't guaranteed to stay in-region |
| jurisdiction | EU only — Scaleway, France (fr-par); one GDPR DPA, no US vendor in the path | United States (Anthropic) | Google Cloud, global infrastructure; no EU-only guarantee |
| who it's built for | People who can't code — describe the question in plain language | Scientists who already use a terminal, SSH, an HPC login node | People who write Python — it's a hosted Jupyter notebook |
| compute power | Browser-side pandas/NumPy/Matplotlib today; heavier EU compute is the Sensitive tier, shipping next | Full HPC orchestration — drafts a plan and submits jobs over SSH/Slurm to your lab's own cluster | Free T4 GPU access, quota-limited and not guaranteed; paid tiers add L4/A100 |
| scientific connectors | Not yet — CSV upload today | 60+ pre-built connectors (UniProt, PDB, Ensembl, ChEMBL, GEO…) plus NVIDIA BioNeMo | Google Drive / Sheets / GitHub integration |
| compliance posture | GDPR-native by construction; tiers map to a university's own data classification | Partial HIPAA; clinical or diagnostic use is explicitly prohibited (beta) | Standard Google Cloud terms; no research-specific compliance framework |
| price | Free to start (€10 in credit); metered after, no seat-based plan required | Bundled into Pro/Max/Team/Enterprise (roughly $17–125+/month) | Free tier (quota-limited); Pro $11.99/mo, Pro+ $49.99/mo |
- Claude Science's provenance and reviewer-agent features — bundling a figure with the exact code, environment, and conversation history that produced it, plus an agent that checks citations and calculations — are real and well-reviewed. MakeMode Notebook shows you the code behind every cell today, but doesn't yet auto-bundle full run provenance or check your work for you; that's on our roadmap, not shipped.
- Claude Science's HPC orchestration (submitting jobs to your own cluster over SSH/Slurm) and its 60+ scientific-database connectors go well beyond a CSV upload — genuinely useful if you already run compute jobs and know your way around a terminal.
- Colab's free GPU access and decade of ecosystem maturity — tutorials, Stack Overflow answers, one-click "Open in Colab" buttons across the ML world — are real advantages if you already write Python.
We're naming this because credibility only comes from conceding the case where the other tool wins. If you already code and run HPC jobs, Claude Science's workbench is a genuinely strong choice. If you have data, a question, and no Python — that's the gap MakeMode Notebook exists to close, on infrastructure your institution can actually approve.
what people are actually asking right now.
Is there a GDPR-compliant alternative to Claude Science?
Does Claude Science work on Windows?
Is there an AI tool for analysing data without knowing how to code, on EU infrastructure?
Does my data leave my computer when I use MakeMode Notebook?
What's the difference between MakeMode Notebook and Google Colab?
run a pilot with your studio course or research group.
We're offering pilot credit to a small number of European research groups and studio courses to try MakeMode Notebook on real data — a first-hand answer to "can I use this on my own dataset, under my own institution's rules?" Small by design: one course, one group, a few weeks.
bring your data, not your code.
No install, no US server in the path, and a strict mode where nothing leaves your device — try it on a real dataset in the next five minutes.