What the agent can do
The agent works on your canvas through the same set of tools that an external AI client would use (see MCP overview). Here's what it can do, grouped by area.
Draw
- Read the current scene to understand what's on the canvas.
- Add, update, and delete elements.
- Add images.
- Create new canvases.
Cards
- List and read cards.
- Place a live card on the canvas.
- Attach a card to an existing element.
- Create a new card.
- Lay out a board as lanes (one lane per list, with cards inside).
- Update a card, comment on it, or toggle its checklist items.
See Card details for how cards behave on the canvas.
Documents
- List, read, attach, and update documents.
- Author reusable MDX components.
See Working with MDX.
Comments
- List, add, reply to, and resolve comment threads.
Library
- Reuse canvas library items and save new ones.
Card-writes safety toggle
By default the agent may freely draw and edit canvas content — elements, images, documents, comments, and canvases. But changing your actual tracker cards is treated more carefully.
Actions that modify real cards — creating a card, updating one, commenting, or toggling checklist items — are OFF by default. To let the agent make those changes in a chat, turn on Allow card writes for that chat. Until you do, the agent will draw and edit the canvas normally but leave your tracker cards untouched.
Turn on Allow card writes only when you want the agent to change the underlying cards, not just the canvas.