marimo — reactive Python notebooks & apps

What it is

marimo is a reactive Python notebook that doubles as an app framework. Open a notebook and every cell that depends on a value automatically re-runs when that value changes — no manual "run all" ordering, no hidden state, no stale outputs. You build interactive dashboards and data apps by mixing Python with marimo.ui components and built-in reactive SQL.

Why it matters

Here is the problem marimo exists to kill: in a traditional notebook, cell output is a screenshot of what happened once, long ago. Edit cell one, and cell twelve still shows the old number until you remember to re-run it. marimo replaces that fragile top-to-bottom ritual with a reactive graph. Each cell is a node; changing an input recomputes exactly the dependent cells. The output on screen is, by construction, the truth.

How it fits your stack

marimo is installed in this environment at version 0.23.9 (the /home/kmail/marimo source checkout). It ships a CLI (edit, run, convert, export, tutorial, pair, new, recover), supports first-class SQL, AI assistance, and deployment as a read-only web app. You can also run a notebook as a plain Python script with arguments.

Example

$ marimo edit notebook.py

  Welcome to marimo!
  Creating a new notebook...
  Serving at http://localhost:2718

  ✓ App served at http://localhost:2718

$ marimo run notebook.py

  Running as a read-only app...
  ✓ App served at http://localhost:2718

  # notebook.py
  import marimo
  app = marimo.App(width="full")

  @app.cell
  def _():
      import altair as alt
      import marimo as mo
      df = mo.sql("SELECT * FROM 'data.parquet'")
      chart = alt.Chart(df).mark_bar().encode(x="month", y="sales")
      mo.ui.altair_chart(chart)
      return chart, df

  if __name__ == "__main__":
      app.run()

$ marimo convert notebook.ipynb

  Converted notebook.ipynb to notebook.py
  ✓ Wrote 5 cells to notebook.py

$ marimo new

  What should the notebook be called? my_app.py
  ✓ Created my_app.py
  
  # reactive example inside
  import marimo as mo
  slider = mo.ui.slider(0, 100, value=50)
  slider  # changing it re-runs dependent cells
  result = mo.md(f"Slider value: **{slider.value}**")

Common flags

edit

create or edit a notebook (opens the browser editor)

run

run a notebook as a read-only web app (deployable)

new

create an empty notebook or scaffold from a template

convert

convert a Jupyter notebook or Markdown file into a marimo notebook

export

export a notebook to HTML, markdown, script, or app

check

check and format marimo files (like a linter/formatter)

tutorial

open a built-in tutorial notebook

recover

recover a notebook from a JSON session file