Canvas and artifacts

Canvas

Canvas is the host's scratchpad for code, scripts and long-form text. When the agent's output reads better as a file than as a chat message (a full script, a refactored class, a long technical document), it opens a canvas in the current conversation.

Opening the canvas in the editor

In the chat header, open the ⋯ menu and choose Open canvas in editor. The canvas opens in a read-only editor tab with VS Code's syntax highlighting. The tab refreshes when the agent changes the canvas while that chat is open in the Verzeta view. To change the canvas, ask the agent, or copy the text into a file of your own. If the chat has no canvas yet, the tab says so and fills in once the agent creates one.

Applying a code block to a file

When a code block in the chat names a file (for example a fence written as ts:src/app.ts or ts path=src/app.ts), it shows the file name and an Apply button. For an existing file, Verzeta opens a diff and asks "apply the highlighted edit to this file?". For a new file, it asks whether to create it. The change is a normal VS Code edit that you can undo. A workspace mount is not needed.

What canvas does not do in the extension

• No editing from VS Code. The canvas tab is read-only.
• No running canvas code from the extension. Canvas Run executes on the host. The extension has no Run button.
• One canvas per conversation. The editor tab shows the conversation's active canvas.

Artifacts

Artifacts are files that an agent or its tools produce while working on a task: generated code files, processed data, reports and similar output. They are stored on the host.

Viewing artifacts in the chat

When a tool produces a file, the artifact appears in the conversation as a row with:

• The file name and type.
• The task or step that produced it, if it came from a multi-step plan.
• The alias of the agent that created it, if there is one.

Accessing artifacts

Artifacts live on the host. In the extension you can see artifact content that appears in chat messages. For large artifacts, or to save them elsewhere, open the Verzeta Studio desktop, which writes the files to disk.

Tip: If your workspace is shared with the chat, the agent can write files straight into your workspace folder as part of a task, so they are already in your project when the task finishes.

Generated media (images and audio)

Images and audio generated by the agent appear in the chat. Generated media in the chat header's ⋯ menu lists them. The host caps each download at 16 MiB. If a generated image is larger, you may see a cut-off preview; get the full file from the desktop, where it was saved.

What these features do not do

• No Run button for canvas code. Code runs on the host.
• No bulk download of artifacts. Use the desktop to manage files.
• No streaming preview of very large files. Open large artifacts on the host.