What is Clusy?
A notebook with an AI agent built in. Describe the work, branch to try other ideas, and keep what holds up.
Clusy is a notebook with an AI agent built into it. You tell the agent what you want (load this file, clean these columns, fit a baseline model) and it writes the code and runs it. You read what comes back, change direction, or take over and type it yourself.
If you've used Jupyter, a lot of this will look familiar. Cells, a running kernel, output under each cell. What's new is the agent sitting next to the notebook, and the fact that you can branch the notebook without recomputing everything you've already loaded.
Clusy runs in the browser at app.clusy.io. There's nothing to install.
Start here
Quickstart
Sign in, ask for something, get a chart back. Start here if you'd rather see it than read about it.
The interface
Chat on one side, notebook on the other — and where everything else lives.
What's actually different
The agent does the work loop
It plans, writes a cell, runs it, reads the output, and decides what's next. It asks you when a call is yours to make.
Branching keeps your state
Fork the notebook at any cell and your loaded data and fitted models come with it. Try a second approach from the same starting point.
Real compute, only when you need it
Everything runs in an isolated sandbox. You attach a GPU for the cells that actually train something, and detach it when they're done.
Your notebooks aren't trapped
Notebooks are .ipynb-compatible. Commit them to GitHub as readable diffs,
download them, or publish to Hugging Face and Kaggle.
Who it's for
People who work with data and models: analysts, data scientists, ML engineers, researchers. You don't have to be one of those, but Clusy assumes you can read Python and a DataFrame and know roughly what you're trying to find. The agent does the typing. You make the calls.
What a session looks like
You describe what you want in chat. There's no notebook yet.
The agent plans, writes the first cells, and runs them. The canvas opens once there's something to show.
You read the output and steer — reply in chat, or edit a cell yourself.
To try something different, you branch instead of overwriting.
You keep what held up: commit it to GitHub, share a read-only link, or publish it.
Run it your way
You aren't locked to Clusy's models. Paste your own Anthropic or OpenAI key and the model bills you directly, on any plan including Free. Or pair a local runner and the turn runs on your own machine, against the Claude or Codex CLI subscription you already pay for.
Where to next
Ready to work? Open app.clusy.io and follow the Quickstart.
Want the concepts first? Read Projects & sessions, then the rest of Core concepts in order.