marimo AI assistant
AI agent · Coding agent · Engineering & development by marimo
AI-native assistance for Python notebooks with context-aware chat, autocompletion, and autonomous cell editing.
The marimo AI assistant is an integrated suite of AI-native features built directly into the marimo notebook editor for developers and data scientists. It provides a range of capabilities from TAB-style code autocompletion and error auto-fixing to a full-featured chat sidebar for generating and refactoring code. The assistant can be configured with different access levels, including a 'Full access' mode where it can read, edit, and run cells independently using a Python scratchpad. It supports a wide variety of models including OpenAI, Anthropic, Gemini, and local models via Ollama, and allows
- Languages
- Python, SQL
- Supported Models
- OpenAI, Anthropic, Gemini, Ollama, OpenAI-compatible APIs
- Access Modes
- Read only, Suggest edits, Full access
- Context Awareness
- Code, outputs, variable previews, and @-tagged variables
Overview
The marimo AI assistant is an integrated suite of AI-native features built directly into the marimo notebook editor for developers and data scientists. It provides a range of capabilities from TAB-style code autocompletion and error auto-fixing to a full-featured chat sidebar for generating and refactoring code. The assistant can be configured with different access levels, including a 'Full access' mode where it can read, edit, and run cells independently using a Python scratchpad. It supports a wide variety of models including OpenAI, Anthropic, Gemini, and local models via Ollama, and allows users to provide specific variable context using @ mentions. By integrating directly with the notebook's reactive runtime, the assistant understands live program state and data schemas to provide highly relevant suggestions.
Notebook interaction
- Read-only access: The assistant can read notebook code, outputs, and variable previews to provide context-aware help.
- Suggest edits: The assistant can propose cell changes for review, allowing users to accept or reject modifications.
- Full access mode: The assistant can read, edit, and run cells directly, with access to a Python scratchpad for independent execution.
- Variable context awareness: Users can @ any variable, dataframe, or database table to provide specific context for AI queries.
Development assistance
- Contextual autocompletion: The assistant provides intelligent code completions that understand project context, imported libraries, and variable types.
- Error auto-fixing: When code encounters an error, the assistant suggests fixes with visual diffs showing proposed changes.
- Integrated chat interface: A dedicated sidebar panel allows users to ask questions, request explanations, or get help with Python concepts.
- Customizable behavior rules: Users can define custom rules in settings, such as preferring specific libraries like Polars over Pandas.
Model and platform support
- Native LLM support: Direct support for models from OpenAI, Anthropic, Google Gemini, and Ollama.
- OpenAI API compatibility: Compatible with any LLM provider that adheres to the OpenAI API standard, including local models.
Specifications
- Supported Models
- OpenAI, Anthropic, Gemini, Ollama, OpenAI-compatible APIs
- Access Modes
- Read only, Suggest edits, Full access
- Context Awareness
- Code, outputs, variable previews, and @-tagged variables
- Interface Options
- Sidebar chat, inline refactoring (Ctrl/Cmd-shift-e), and TAB completion
Added by AntSpace from public sources (marimo.io).