marimo

marimo is a revolutionary open-source Python notebook that reimagines interactive computing for the modern era of data science and artificial intelligence. Unlike traditional notebooks that store content in error-prone JSON files, marimo notebooks are pure Python scripts (.py files). This fundamental design choice makes them reproducible, Git-friendly, testable, and reusable as standard Python modules. marimo introduces a reactive execution model, where updating a single cell automatically triggers the execution of all dependent cells, eliminating hidden state and the tedious manual task of re-running code to ensure consistency. This creates a fluid, intuitive, and error-resistant development environment.

Developed by a team with extensive experience from Google, Palantir, and the PyData ecosystem, marimo is built to bridge the gap between experimental prototyping and production-ready applications. It empowers developers and data scientists to build complex, interactive web applications directly from their notebooks without writing any frontend code. These notebooks can be seamlessly shared as interactive HTML files powered by WebAssembly or deployed as robust web apps using the marimo command-line interface.

How to use marimo

Getting started with marimo is straightforward and designed for a smooth developer experience. First, install it using pip:

$ pip install marimo

Once installed, you can launch the introductory tutorial to familiarize yourself with its reactive paradigm:

$ marimo tutorial intro

To create or edit a notebook, simply run the marimo command with a Python file:

$ marimo edit my_notebook.py

Inside the web-based editor, you can write Python code in cells, just like in other notebooks. However, marimo's power lies in its special features. You can bind UI elements like sliders and dropdowns directly to Python variables. When a user interacts with a widget, the corresponding variable updates, and all downstream cells that use this variable reactively re-run. You can also write SQL queries directly in dedicated SQL cells to interact with databases like DuckDB, SQLite, Postgres, and more. To share your work, you can either run it as a read-only application or export it to a self-contained HTML file:

# Run as an interactive app
$ marimo run my_notebook.py

# Export to HTML
$ marimo export my_notebook.py --to html

The integrated AI assistant can be used to generate code, explain concepts, and debug issues, supporting models from OpenAI, Anthropic, Google Gemini, and local models via Ollama.

Core Features of marimo

• Reactive Execution: Code and outputs stay perfectly in sync. Modifying a cell automatically updates all dependent cells, eliminating hidden state and ensuring reproducibility.
• Git-Friendly Format: Notebooks are saved as standard Python (.py) files, not complex JSON. This results in clean, readable diffs, making version control with Git seamless.
• AI-Native Editor: Features integrated AI assistance, including GitHub Copilot-style autocomplete, context-aware chat, and inline edits. It supports major AI models to accelerate development.
• Built-in SQL Cells: Natively query databases and dataframes using SQL. The results can be seamlessly transformed and visualized using the full power of Python.
• Interactive UI Elements: Build interactive UIs with a rich set of widgets like sliders, dropdowns, and interactive dataframe GUIs without needing any frontend expertise.
• Seamless Deployment: Effortlessly transition from a prototype notebook to a production-grade tool. Deploy notebooks as read-only web apps, run them as Python scripts in pipelines, or export them as self-contained HTML files.
• Reproducible Environments: marimo can serialize package requirements within the notebook file and automatically create ephemeral virtual environments, ensuring that notebooks run consistently everywhere.
• Rich Developer Experience: The editor includes modern features like vim keybindings, code formatting, debugging panels, extensive hotkeys, and a VS Code extension.

Use Cases for marimo

marimo's versatility makes it suitable for a wide range of applications across different domains:

• AI and Machine Learning Development: Rapidly prototype and test AI models. Build interactive interfaces for LLMs, such as chatbots and generative UI tools. Create custom data labeling and evaluation dashboards to fine-tune and compare models.
• Data Exploration and Visualization: Explore datasets interactively using reactive plots and dataframe GUIs. Selections made in a plot can automatically update downstream analyses and visualizations, creating a dynamic and exploratory workflow.
• Scientific Research and Education: Communicate research findings through tangible, interactive applications. Develop engaging educational materials that allow students to experiment with concepts in real-time, such as simulating physical processes or exploring mathematical functions.
• Internal Tooling and Dashboards: Build and maintain internal tools like financial calculators, NLP model comparison dashboards, or data labeling interfaces using only Python. This eliminates the complexity of managing separate frontend and backend services.

Advantages of marimo

marimo offers significant advantages over traditional notebook environments like Jupyter:

• Reproducibility by Design: The reactive, acyclic dataflow graph guarantees that notebooks execute in a deterministic order, preventing the common "out-of-order execution" problems that plague other systems.
• Simplified Version Control: By using a pure Python format, marimo solves the nightmare of resolving merge conflicts in notebook JSON files, enabling true collaboration via Git.
• From Notebook to App, Instantly: marimo unifies the development and deployment process. What you build in the notebook is the app, drastically reducing the time and effort needed to share your work.
• Modern and Integrated Toolchain: It combines the best of modern development practices, including built-in AI assistance, first-class SQL support, and powerful interactive components, all within a single, cohesive environment.
• Open Source and Community-Driven: As a free and open-source tool with a vibrant community on GitHub and Discord, marimo is continuously evolving with contributions and feedback from its users.

Pricing and Plans

marimo is fundamentally a free and open-source project. The core marimo notebook library and its command-line interface are available for anyone to use without charge. You can install and run it on your local machine or your own servers. Additionally, the project offers molab, a cloud-based platform to run and share marimo notebooks, which is also currently available for free. This positions marimo as a highly accessible tool for individuals, students, and professional teams alike.