Let your agent see what you see.
Point to a chart, table, or other result in marimo, a reactive Python notebook. Lens gives your agent the selection, your note, and the code behind it, then brings the agent's work back into view for your review.
https://github.com/user-attachments/assets/e1975ba6-3087-485e-9360-43a99adc8795
- Point and ask. Mark a point or region on a chart, table, or other output. Lens keeps the mark, note, image, and relevant notebook context together.
- Follow the work. See where the agent is working, or ask for a guided trail through the notebook's inputs and results.
- Review and continue. Reopen an addressed selection from History to refine the request against the current output.
Get started
Use Lens with an agent connected to your notebook through marimo Pair or marimo's code-mode sidebar. Lens is a separate Python package, installed in the notebook's environment.
Follow the quickstart to connect an agent, select a chart, and review your first change, or open the sample notebook in molab:
For agents
Read the packaged Lens briefing:
uvx --with marimo-lens agent-plugins read marimo-lens
In a connected notebook kernel, use import agent_plugins as ap followed by
print(ap.read("marimo-lens")) to read that environment's installed guidance.
Connect an agent covers
connection and resource access.
Documentation
What is Lens? · Getting started · Try the demo · Connect an agent · Python API
Contributing · Report a bug · Apache 2.0
Inspired by Agentation.