Using your own editor

While we recommend using the marimo editor, we understand that you may prefer to use your own. marimo provides a --watch flag that watches your notebook file for changes, syncing them to the marimo editor or running application. This lets you edit your notebook using an editor of your choice, like neovim, VSCode, Cursor, or PyCharm, and have the changes automatically reflected in your browser.

Install watchdog for better file watching

For better performance, install watchdog. Without watchdog, marimo resorts to polling.

marimo's file format

File format tutorial

Run marimo tutorial fileformat at the command line for a full guide.

marimo stores notebooks as Python files. Cells are stored as functions, decorated with@app.cell; you can optionally give cells names in the editor UI or by editing the notebook file.

@app.cell
def memorable_cell_name(auto, determined, references):  # signature denotes cell references
    computed_value = auto + determined + references
    "hello!"                                            # final statement is the visual output
    return computed_value                               # return denotes cell definitions

Cell signature and returns

Don't worry about maintaining the signatures of cells and their return values; marimo will handle this for you.

Exposing functions and classes top-level

You can expose top-level functions and classes in your notebook, so that other Python modules can import them:

from my_notebook import my_function, MyClass

Top-level functions are added to a notebook using the @app.function decorator, and classes with @app.class_definition; these appear in your notebook as cells with just a function or class definition. These functions and classes must be pure, closing over only other pure functions and classes, or imports and constants defined in an app.setup with block.

Here is a complete example that you can copy/paste and run locally:

import marimo

app = marimo.App()

with app.setup:
    # These symbols can be used by top-level functions and classes
    # (as well as by regular cells)
    import numpy as np

    CONSTANT: int = 1

@app.function
def my_function(x: np.ndarray):
    return np.mean(x) + CONSTANT

@app.class_definition
class MyClass:
    ...

@app.cell
def _():
    my_function(np.random.randn(2, 2))
    return

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

For more details see the guide on reusable functions and classes.

Types and autocompletion

Add type hints to your variables, and marimo will carry over these type hints to cells where these variables are used. This, combined with importing modules in the setup cell (see below for an example), makes it possible for your editor to give completions on the references of your cell.

For example:

# setup cell
import numpy as np

# cell 1
x: np.ndarray

# cell 2
np.mean(x)

will be serialized as

import marimo

app = marimo.App()

with app.setup:
    import numpy as np

@app.cell
def _():
    x: np.ndarray
    return x,

@app.cell
def _(x: np.ndarray):
    np.mean(x)

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

As markdown

Markdown File format tutorial

Run marimo tutorial markdown-format at the command line for a full guide.

marimo notebooks can also be stored as Markdown files. This is a good option for prose heavy text, and can be easy to navigate and edit in external editors. To convert a marimo notebook to markdown, use

marimo export md notebook.py -o notebook.md

at the command-line, or rename your file to have an .md extension in the notebook editor.

marimo conforms to standard markdown document format, and will render most places like Github. Metadata in this file format is saved in the frontmatter, which marimo may use for information like sandboxing, and the marimo version. All other fields are kept, but ignored.

For execution, marimo extracts code fences that contain marimo in braces. For instance python {marimo}, {marimo} or {.marimo .python}. The marimo editor uses python {.marimo} which is Pandoc compatible, and correctly processed by text highlighters.

---
title: My Notebook
marimo-version: 0.0.0
description: A notebook with a description
---

# Just a notebook

```python {.marimo}
print("Hello World!")
```

marimo's markdown format can be used with a mkdocs plugin and Quarto.

Note that the markdown format is not as fully featured as the Python format. Reactive tests will not work, markdown notebooks cannot be imported or used as a library, and they cannot be run as scripts.

Watching for changes to your notebook

marimo edit --watch

When you run marimo edit with the --watch flag, the marimo server will open your notebook in the browser and watch the underlying notebook file for changes. When you make changes to the notebook file, they will be streamed to the marimo editor in the browser.

By default, synced code will not be executed automatically, with cells marked as stale instead. Run all stale cells with the marimo editor's "Run" button, or the runStale hotkey, to see the new outputs.

If you want to run all affected cells automatically when you save, change the runtime config in your pyproject.toml file.

[tool.marimo.runtime]
watcher_on_save = "autorun"

marimo run --watch

When you run marimo run with the --watch flag, whenever the file watcher detects a change to the notebook file, the application will be refreshed. The browser will trigger a page refresh to ensure your notebook starts from a fresh state.

Watching for changes to other modules

marimo can also watch for changes to Python modules that your notebook imports, letting you edit auxiliary Python files in your own editor as well. Learn how to enable this feature in our Module Autoreloading Guide

Watching for data changes

Note

Support for watching data files and automatically refreshing cells that depend on them is not yet supported. Follow along at https://github.com/marimo-team/marimo/issues/3258 and let us know if it is important to you.

Hot-reloading WebAssembly notebooks

Follow these steps to develop a notebook using your own editor while previewing it as a WebAssembly notebook in the browser. This lets you take advantage of local development tools while seeing the notebook as it appears when deployed as a WebAssembly notebook.

# in one terminal, start a watched edit (or run) session
marimo edit notebook.py --watch

# in another terminal
marimo export html-wasm notebook.py -o output_dir --watch

# in a third terminal, serve the WASM application
cd path/to/output_dir
python -m http.server  # or a server that watches for changes