Skip to content
← Tools / Skills
Toolmarimo/10 trustVerified
marimo logo

marimo

Reactive Python notebooks stored as pure .py — AI-native, git-friendly, no hidden state.

Acquired — still operating

acquired by CoreWeave Oct 2025

Visit official site →

About

marimo is an open-source reactive Python notebook (Apache 2.0, roughly 20,000–22,000 GitHub stars) that stores notebooks as plain .py files instead of JSON, executes cells in a dependency graph so downstream cells update automatically, and eliminates hidden state for reproducible, pytest-testable notebooks. It bundles first-class SQL cells (DuckDB, PostgreSQL, MySQL, SQLite, Snowflake), interactive widgets without callbacks, marimo run to deploy notebooks as web apps, WASM export, and an AI-native editor with built-in assistants plus marimo pair, which connects agents like Claude Code to live sessions.

In October 2025 marimo was acquired by AI cloud company CoreWeave, and it is used by research teams at Stanford, Berkeley, and SLAC as well as AI engineering teams. Notebooks run interactively, as CLI scripts, as apps, or in the browser via Pyodide.

Limitation: in 2026 a critical pre-authentication RCE (CVE-2026-39987, severity 9.3) in the terminal WebSocket endpoint was exploited in the wild within 10 hours of disclosure — never expose a marimo server to the internet, and upgrade past 0.23.0. It also trades Jupyter's huge extension catalog for cleaner diffs, so teams deep in the Jupyter ecosystem should pilot first.

What it can do

  • Reactive cells with automatic dependency-graph re-execution
  • Notebooks stored as pure .py — git diffs, pytest, script execution
  • First-class SQL cells across DuckDB, Postgres, MySQL, SQLite, Snowflake
  • AI-native editor with agent pairing and app/WASM deployment

Setup

Install with pip install marimo, then marimo edit to open the editor; or run notebooks in the browser via molab.

Prerequisites

Python 3.8+; a browser for the editor UI.

Use cases

data-sciencenotebookssqlprototyping

Works with

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 61; Trending score 82/100 from researcher assessment.

★ 22,000 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.