Summary

marimo-pair enables AI agents to interact directly with running Marimo notebook sessions, transforming reactive Python notebooks into dynamic environments for agent execution. It targets users seeking to automate tasks or build agent-driven workflows within the Marimo ecosystem, offering a novel way to leverage notebooks as agent backends.

How It Works

The project integrates agents into Marimo notebooks by exposing notebook functionalities as an environment. It leverages the "Agent Skills" open standard for seamless integration with compatible agents. The core mechanism involves agents discovering and interacting with running Marimo servers, allowing them to execute code and manipulate notebook states programmatically.

Quick Start & Requirements

• Installation: Use npx skills add marimo-team/marimo-pair or uvx deno -A npm:skills add marimo-team/marimo-pair. For Claude Code, add the marketplace (/plugin marketplace add marimo-team/marimo-pair) and install the plugin (/plugin install marimo-pair@marimo-team-marimo-pair).
• Prerequisites: A running Marimo notebook (configured with --no-token or MARIMO_TOKEN), bash, curl, and jq available on PATH.
• Setup: Integration is typically quick, dependent on having Marimo and the necessary CLI tools installed.
• Docs: No direct links to quick-start or demo pages were provided in the README excerpt.

Highlighted Details

• Supports any agent adhering to the Agent Skills open standard.
• Offers auto-discovery for running Marimo notebook servers.
• Provides a dedicated plugin for Claude Code, enhancing agent capabilities within that environment.
• Includes mechanisms to manage bash command execution permissions for security-conscious environments like Claude Code.

Maintenance & Community

No specific details regarding contributors, community channels (like Discord/Slack), or roadmap were provided in the README excerpt.

Licensing & Compatibility

The README excerpt does not specify the project's license or provide compatibility notes for commercial use or closed-source linking.

Limitations & Caveats

When using with Claude Code, users may encounter prompts to approve bash commands; this can be mitigated by explicitly adding script paths to ~/.claude/settings.json or project-level settings. The system relies on external command-line tools (bash, curl, jq) being present and executable.