Python notebooks with AI

marimo is modern Python notebook. So that means ample integrations with AI tools. You have access to cutting edge coding agents, which in turn have access to a rich development environment that can benefit from interactive widgets as well as database connections.

AI integration for Python notebooks

marimo offers a few ways to integrate AI into your workflow. You can configure TAB-style autocompletion, you can autofix errors in a cell with rich diffs or you can use the chat interface in the sidebar.

• The TAB autocompletion goes beyond simple syntax suggestions. It understands your project context, imported libraries, and variable types to provide intelligent code completions that actually make sense for what you’re trying to accomplish. Instead of generic suggestions, you get contextually relevant code that fits naturally into your workflow.
• The error auto-fixing feature is particularly powerful. When your code encounters an error, marimo can automatically suggest fixes with visual diffs showing exactly what would change. This isn’t just about syntax errors - the AI understands logical issues, import problems, and even suggests more efficient implementations.
• The integrated chat interface lets you ask questions about your code, request explanations, or get help with specific Python concepts without leaving your notebook. You can highlight code and ask for optimizations, request documentation, or even ask for help implementing new features.

You can also add custom rules in your user settings, which allows you to make the LLM aware of your preferences. For example prefer polars over pandas is a common line.

Connect coding agents with marimo pair

Beyond the in-editor assistant, marimo lets full coding agents work inside a running notebook. marimo pair is an agent skill that drops your agent of choice into a live notebook session, giving it access to everything that you see. The agent can run code in an ephemeral scratchpad, add and delete cells, install packages, and inspect the notebook’s in-memory variables and state. The notebook becomes a shared canvas for you and the agent, as well as a reactive REPL that eliminates hidden state and guarantees a reproducible Python program. You can use marimo-pair with any agentic harness that can use skills.

You can find marimo pair on GitHub. If you’d rather drive an agent from inside the editor, marimo’s built-in Agent sidebar connects ACP-compatible agents directly to your notebook so they can write and modify cells as you steer.

Support for many LLMs

marimo has direct support for LLM models from OpenAI, Anthropic, Google Gemini and Ollama. However, marimo also supports any LLM that can understand the OpenAI API standard. So you’re free to experiment with any provider that you like, including those that run locally on your machine.

The OpenAI API compatibility also opens up access to newer providers and open-source models as they become available, and marimo ships out-of-the-box support for the latest frontier models as they land. You can also plug in gateways built for coding agents — like OpenCode Zen, a curated model gateway, or OpenCode Go, a low-cost subscription to open models like Kimi. You’re not locked into any single vendor’s ecosystem, giving you the freedom to adapt as the AI landscape evolves.

And if you want to try AI without configuring your own keys, molab — marimo’s free, cloud-hosted notebook — comes with built-in AI assistance out of the box.

AI that is aware of your code

You can @ any variable that marimo has access to in order to add context. This includes dataframes as well as any tables in a connected database. This really helps with writing SQL as well as debugging.

As you can see, this also brings up some much appreciated autocompletion.

Want to give marimo a spin?

uv pip install marimo