Skip to main content
Canvas gives you a dedicated workspace beside chat. When a reply becomes something you may want to keep working on, Babbily can open it as an artifact in the Canvas panel. Everything you make is also collected on the Artifacts page in the sidebar. Use artifacts for documents, tables, code, and interactive visuals that need more than a one-time chat reply.

Chat and models

Learn how Canvas fits into regular chat threads.

Visual replies in chat

See the visuals Babbily can show in a reply and save as artifacts.

When Canvas appears

Babbily may open Canvas when you ask for substantial work that benefits from a separate editing surface. Artifacts can be:
  • Documents, such as drafts, notes, and Deep Research reports.
  • Code, such as snippets and scripts.
  • Sheets, for tables and spreadsheet-style data.
  • Visuals, such as interactive apps, tools, mockups, dashboards, and diagrams.
If a response includes an artifact card, open it to continue working in Canvas. The model you chose, or the one Orchestrator picked, decides what a document, code file, or sheet should contain. GPT-6.1 Terra, a fast writing model, then writes it, whichever model you use, and the writing counts toward your API usage budget at GPT-6.1 Terra’s rate. Visual artifacts are drawn separately.

Make an artifact on purpose

To make sure a reply becomes an artifact, choose a create command from the composer or type it: You can also ask in plain words. Asking for an app, a tool, or an interactive mockup makes a visual artifact.

Find everything in Artifacts

Select Artifacts in the sidebar to see every artifact in the active workspace, newest first.
  • Filter by kind. Choose All, Visuals, Documents, Code, Sheets, or Images.
  • Search by name with Search artifacts.
  • Sort by Recent activity, Oldest first, Name A–Z, or Name Z–A.
  • Switch layouts with Grid view and List view.
  • See who can read it. Each card shows when it was last edited and who else can see it, such as Private, Published with its view count, Shared with team, In a shared chat, In a public chat, or In a team project.
Visual artifacts show a picture of the visual on their card, so you can recognize them at a glance. In a team workspace, tabs let you switch between All, Yours, and Shared with you. Artifacts a teammate shared show who shared them.
Artifacts used to appear in Library > Files & Artifacts. They now have their own page, and Library keeps your chats, media, and translations.

Make something new

The top of the Artifacts page has starters for the most common kinds of work: Choose a starter to open a new chat with the request already started in the composer. Finish describing what you want, add any files or data, and send it.

Open an artifact

Select a card to open the artifact on its own page. From there you can:
  • Step through earlier and later versions.
  • Select Open in chat to return to the chat that made it.
  • Select Share to share or publish it.
  • Select Export to download it in the formats it supports.
Each card on the Artifacts page has the same Export menu behind its download icon.

Share an artifact

You can share an artifact without sharing the whole chat. Open it from the Artifacts page and select Share to open Share artifact.
  • Publish a public link. Turn on Make link public to get a Public link. Anyone with the link can see the published version, signed in or not.
  • Keep the link current. A published link shows the version you published. When you save a newer version, choose Update to latest version to publish it.
  • See views. The dialog and the artifact’s card show how many views the public link has.
  • Share with your team. In a team workspace, share the artifact with Everyone in the team or with specific teammates. They find it under Shared with you on their Artifacts page and can view it but not edit it.
  • Stop sharing. Turn the public link switch off to take the link down, or change the team sharing in the same dialog.
An artifact can be published once it’s finished. If it is still being written or waiting on an approval, the switch stays off until it completes.
Review an artifact before you publish it. Anyone with the public link can see the published version, including any data you put in it.

Review a Deep Research report

When you use Deep research, Canvas opens after the research plan is ready. The artifact can show:
  • The questions the agent is working through.
  • The planned report sections.
  • Research progress and sources found so far.
  • The cited report as it is written.
The artifact card updates when the run finishes. If evidence gathering reaches its time limit, Babbily finishes from the available sources and notes in chat that coverage is partial. If writing reaches its limit, Canvas keeps the written portion and labels the report incomplete. If you selected Plugins, Deep Research reads from them while it gathers evidence. It doesn’t send, create, or update anything in a connected app, and it doesn’t pause for approval.

Deep research and search

Learn when to use the Deep Research Agent, how to follow its plan, and how to review sources.

Work with visual artifacts

A visual artifact is an interactive page Babbily draws for you, such as a working calculator, a small app, a dashboard, a mockup, or a 3D scene. It opens in the Canvas panel and on its own page, where you can use it the same way you use a visual in a reply.
  • Make one with Create visual, the Design or Dashboard starter, or a request for an app, tool, or interactive mockup.
  • Save one from a reply. Open a visual’s … menu in a chat and choose Save as artifact.
  • Change it by asking. Describe what to change in the same chat, such as “add a dark mode toggle” or “show the totals as a bar chart.” Babbily draws a new version.
Visual artifacts run in their own isolated space, separate from your account, just like visuals in replies. See visuals and your privacy. Older Generated UI artifacts now open as a readable text outline. You can still download them as JSON.

The Canvas panel

On a computer, the panel sits against the right edge of the window, the full height of the page, with a thin line between it and the chat. Drag that line to resize the panel. Its header is one row: the artifact’s title, then Save when you have unsaved changes, Download for code, Export, and Close. In a narrow window, the header keeps the title and shows icons for the buttons. On a phone, the panel opens full screen.

Edit directly in Canvas

Canvas lets you edit supported artifacts without copying them into another editor first. Depending on the artifact type, you can:
  • Edit text or markdown.
  • Adjust table-like content.
  • Review code in a focused editor.
  • Rename the artifact.
  • Save a new version.
  • Resize the side panel on a computer.
  • Close the panel and reopen the artifact later.
Visual artifacts aren’t edited by hand. Ask Babbily for changes instead. Save appears in the header once you’ve changed something, and saves your edits as a new version. An artifact with no unsaved changes shows no Save. While Babbily is still writing an artifact, Save is unavailable. Tables, code blocks, and the rest of a document’s formatting stay as they are when you edit one part of it and save.
Save your changes before switching away if you want Babbily to keep that version.

Revise with the latest Canvas state

When you continue prompting in the same thread, Babbily can use the active Canvas artifact as context. That helps follow-up requests apply to the current version, including edits you made in the panel. Good follow-up prompts include:
  • Make this shorter and keep the same section order.
  • Turn the second section into a table.
  • Update the code to accept a CSV file.
  • Add a warning callout under the pricing notes.
If you want to change topics, start a new thread or close the artifact so old Canvas context does not steer the next request. If you use Model Orchestrator in the model picker, Orchestrator uses its dependable default model while an artifact is open, so the artifact’s content is taken into account.

Use versions

Saving changes creates a new artifact version when version history is available. An artifact with more than one version shows a Versions row under the panel’s header, such as Versions ‹ 2/3 ›. Use its arrows to move between earlier and later versions. Versions are useful when you want to compare a draft before and after a prompt, recover older wording, or review how a table changed.

Export and download

Select Export in the panel’s header, then choose a format. Each format is listed with what it’s for, such as PDF, “A document to print or share”. Babbily downloads the file and shows the export running on the Export button. Code also has its own Download button. Export options depend on the artifact type. Some visuals can’t be saved as a picture, and 3D scenes can’t be saved as a page. Those options are hidden when they don’t apply. Review exported files before sharing them outside Babbily.

Send artifacts to connected apps

Some Canvas artifacts can be sent to connected apps from the Export menu, such as Google Docs, listed under the download formats. Its row shows the send’s progress as it checks, sends, and finishes. Available destinations depend on your connected apps, permissions, and whether the plugin tools are ready. If a destination is not connected, Babbily may ask you to connect it first. If an action writes to another app, review what will be sent before approving it. Deleting the chat affects sends like this:
  • If you delete the chat before a pending send to Google Docs starts, the artifact isn’t sent.
  • Deleting the chat doesn’t delete a document that was already sent. Remove it in Google Docs.

Plugins

Connect workspace apps and review approval-required actions.

Run code artifacts

When a code artifact supports execution, use the run action in Canvas to test it. Code execution is intended for quick checks, not production workloads. Execution support can vary by language and artifact state. JavaScript, TypeScript, JSX, TSX, and Python-like code can run through supported JavaScript or Python execution paths. Network access is unavailable during execution, and long-running code may stop before it finishes.
Review code before running it. Do not include passwords, API keys, private keys, customer data, or other secrets in code artifacts.

Limits to expect

Canvas availability can vary by model, plan, tool mode, and the type of work you request. Keep these expectations in mind:
  • Babbily may answer in chat instead of Canvas when the request is simple.
  • Visual artifacts take a little longer, because Babbily draws them, and use more of your API usage budget than a plain reply.
  • Export and plugin actions vary by artifact type.
  • Plugin actions depend on your app permissions and plugin readiness.
  • If writing an artifact stops partway, for example because it ran out of time, the artifact says This artifact took too long to generate. above what was written. Ask again to get a complete version.
  • If you stop a reply while Babbily is writing an artifact, its card appears in the thread right away with what was written so far.
If Canvas does not appear, choose Create document, Create visual, or Create sheet, or ask directly for a draft, table, code file, or interactive visual.
Last modified on November 21, 2026