marimo is a reactive Python notebook designed to address fundamental limitations of traditional notebook environments like Jupyter. Stored as pure Python files, marimo notebooks are reproducible, version-controllable with git, executable as scripts, and deployable as interactive web applications. The project positions itself as a batteries-included replacement for multiple tools including Jupyter, Streamlit, jupytext, ipywidgets, and Papermill.
The core innovation in marimo is its reactive execution model. When a user runs a cell or interacts with a UI element, marimo automatically executes all dependent cells to maintain consistency between code and outputs. This eliminates the hidden state problems endemic to traditional notebooks where cells can be executed out of order, leaving the program in an inconsistent state. The system can also be configured to mark affected cells as stale rather than automatically executing them, accommodating expensive computations while still maintaining guarantees about program state.
marimo includes comprehensive data handling capabilities with first-class SQL support. Users can write SQL queries that depend on Python variables and execute them against dataframes, databases, lakehouses, CSVs, Google Sheets, and other data sources. The built-in SQL engine returns results as Python dataframes, keeping notebooks pure Python despite SQL integration. Interactive dataframe features allow users to page through, search, filter, and sort millions of rows without writing code.
The editor is modern and AI-native, supporting GitHub Copilot, AI assistants specialized for data work, vim keybindings, a variable explorer, and code completion. marimo can be used through its web-based editor, VS Code via an official extension, or any text editor with file watching. The system includes built-in package management supporting all major package managers, with the ability to install packages on import and serialize requirements directly in notebook files.
According to GitGenius activity tracking, the repository shows strong engagement with a median issue and pull request response latency of 0.0 hours and a mean latency of 112.6 hours across 2760 tracked items. Bug reports represent the most active issue category with 1529 items, followed by enhancement requests with 807 items. The project maintains active discussion with 129 items tagged as needing discussion. Primary contributors mscolnick, akshayka, and dmadisetti have driven 4830, 1248, and 825 events respectively, indicating concentrated but sustained development effort.
The repository overlaps with contributors from github/gh-aw, solo-io/gloo, and microsoft/vscode, suggesting cross-pollination with command-line tools, API gateway projects, and the VS Code ecosystem. GitGenius classification places marimo across diverse domains including machine learning, data analysis, interactive notebooks, real-time analytics, computational tools, scientific computing, and event-driven architecture, reflecting its broad applicability across data science and research workflows.
marimo notebooks are executable as Python scripts parameterized by command-line arguments, deployable as interactive web apps or slides, and runnable in browsers via WebAssembly. Functions and classes can be imported from one notebook to another, enabling code reuse. The system supports pytest testing of notebooks and includes dynamic markdown parametrized by Python variables for narrative-driven data exploration. Built-in features encompass HTML export, fast code completion, and an interactive dataframe viewer, positioning marimo as a comprehensive environment for reproducible research, data analysis, and interactive application development.
