marimo-studio turns one marimo notebook into reports,
apps, and presentations. Keep data, computation, and controls in Python.
Shape each view for its audience, by hand or with a coding agent.
Experimental: Studio is changing rapidly. Pin
marimo-studioin saved projects.
Open a notebook with Studio installed:
uvx --with marimo-studio marimo edit analysis.py --sandboxAdd a displayable cell, then click Add view in the Studio toolbar. Save the notebook if prompted, choose HTML document, and create the view. Notebook and Preview open side by side. The source action opens the view's files, and marimo's agent sidebar stays available.
Saving view source rebuilds Preview. Notebook controls keep their reactive behavior, and a failed build retains the last successful view.
Follow the quickstart
for a complete first view. uv supplies uvx and
resolves the notebook's declared dependencies with --sandbox.
The Rio 2016 notebook supplies one analysis to three views:
| View | Explore |
|---|---|
| Overview | Roster totals, delegations, sports, and medalists |
| Explorer | Filter the roster and brush linked charts |
| Field | Move through the athlete data in a four-chapter presentation |
Build views with HTML, React, Svelte, or Observable Notebook Kit. Each view owns its source and browser dependencies. Explore all examples.
Give a terminal agent such as Claude Code or Codex one instruction:
claude 'Follow `uvx --with marimo-studio agent-plugins read marimo-studio`
to build a briefing view of analysis.py that leads with the headline results.'agent-plugins read prints the briefing Studio ships for coding agents through
Agent Plugins. It tells the agent
how to pair with your running notebook, or start one, then create, build, and
show the view in Preview beside the notebook. In marimo's AI sidebar in
Code Mode, ask for the view directly.
The agent guide covers writing a good request and sending visual feedback with Lens.
Use a live Python server, execute Python in the Browser, or export Prepared results as a static site. Prepared delivery publishes the exported results and input states while keeping Python source on the build machine.
Run or export a view covers the delivery choice and what visitors receive.
Guide 路 Reference 路 Troubleshooting 路 Security 路 Contributing