Cursor Agent: The Canvas Tool

Canvases are interactive artifacts that Cursor renders next to the chat. Instead of scrolling through a long markdown table or code block, you get a standalone view laid out with sections, stats and tables, which you can reopen, edit and iterate on. When you ask an agent for a dashboard, analysis, audit or report, Cursor opens the result in a canvas if that fits better than plain chat output.

Whether you want a canvas depends on what happens to the output after you read it. Chat output gets scrolled past. A canvas is saved to your workspace's canvas list, so you can reopen it and rerun it with fresh data.

Ask whether anyone will want these numbers again in a month. If nobody will, the long markdown table in the chat was already fine.

The loop has four steps, from the trigger to a canvas you can rerun later.

1. 1Cursor decides your task benefits from a visual or interactive view, or you ask for one directly.
2. 2Cursor builds the canvas and inserts a reference to it in your chat.
3. 3You review the rendered view, switch to the source to tweak it, or ask Cursor to change it.
4. 4Cursor saves the canvas so you can reopen and rerun it later with fresh data.

Each canvas appears in your workspace's canvas list, so you can jump back to past ones without rerunning them.

Cursor drives most of that loop, but step three is yours, and it decides whether the canvas is worth reopening. A canvas built on a query with the wrong date range renders exactly as well as one built on the right range. Switch to the source once before you move on.

If you wanted a canvas and none appeared, ask for one. Step one fires on either trigger, so you never have to wait for Cursor to decide the task deserves a canvas.

Cursor gives you three ways in. Only the first assumes you still have the chat that produced the canvas open.

• Click the card that appears at the end of the response when Cursor creates a canvas.
• Run Open Canvas from the Command Palette, where it is listed under View.
• Open a canvas tab from the new tab menu in the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition..

The card is the quickest route while that chat is in front of you. For work you come back to later, the palette command and the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. tab don't depend on finding the chat again.

Cursor's docs describe the canvas list as your workspace's, and the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. as the place to work with agents across all your projects. They don't say whether a canvas tab opened in the Agents Window reaches past the workspace you are in, so test it before you plan a workflow around it.

Sharing gives teammates a link to a live snapshot of the view, with the same layout, charts and tables. They open it in the browser without rerunning the agent or digging through chat history. Use Publish on the canvas toolbar to publish a share or refresh one. The Shared Canvases page on the dashboard lists the canvases you published yourself, not your teammates', so ask a teammate for the link to theirs.

A share is a snapshot, and the same Publish button both creates it and refreshes it. Cursor's docs don't say the link updates itself, so don't promise anyone that it will. For a view people need fresh every week, the other route is a canvas skill committed with the code, which lets a teammate regenerate the view from a short prompt. Decide which one your audience needs before you publish.

A teammate can open your canvas through one documented route: a share link you publish. Until you publish, the canvas sits in your workspace's canvas list.

Once published, a link has no documented expiry. Cursor's enterprise hardening page lists share links for chats and canvases under data retention and deletion, and points to the dashboard to delete them. Keeping the published set small is our habit, not a rule Cursor states. Retire a share when the analysis behind it has been superseded, before a teammate notices the number looks wrong.

Canvases are built to be refined, and Cursor's guidance is to tell the agent what to change before you reach for the source.

• If the layout isn't right, tell Cursor what to change instead of editing by hand.
• If the numbers look stale or off, ask Cursor to rerun the underlying query or show its work.
• For larger reworks, revert and prompt Cursor again with more details. This is usually faster than nudging through small follow-ups.
• For small tweaks, you can also edit the source code by hand.

Editing the source looks like the developer's move, since the layout is right there. Cursor still puts describing first and keeps hand edits for small tweaks, and that order pays off later. A sentence that describes the change can go into a skill's layout instructions when you need the same report again. A hand edit doesn't carry over to the next canvas.

Design ModeA way to point at an element in Cursor's built-in browser and change it directly, instead of describing it in words. Press Enter for the full definition. also runs inside canvases, so you can select and annotate the element you want changed instead of describing it in prose. Canvas tabs keep the older pen-icon toggle, and notes there go straight to the chat. The Design Mode guide covers how it works.

You can package a common canvas workflow as a skill, so Cursor produces a consistent layout every time you ask. Cursor's docs list four parts a canvas skill typically includes.

Start with the trigger description, because it decides whether the agent reaches for the skill on its own. The agent uses a skill's description to judge whether the skill is relevant, so a vague trigger can leave even careful layout instructions unused. When a canvas skill isn't firing, rewrite the description before you touch the layout.

Once the skill is in place, a short prompt regenerates the canvas with fresh data, and every teammate using the skill gets the same output shape.