Artifacts turns Claude from a chat AI into a workshop. Code, documents, HTML pages, diagrams — they live in a side panel where you iterate, refine, and ship. Most people never realize when to invoke it, or what the real win is. This lesson shows it in action: when Artifacts wins, when it gets in the way, and the iterate-in-place loop that makes Claude a building partner.
Update · November 10, 2026
Artifacts now come in team-sized types: Whiteboards, Sheets and Data
Artifacts are no longer just a side panel for code and pages. Anthropic added three shared, editable artifact types that sit alongside the docs, slide decks and designs Claude can already make. All three are on every paid plan (Pro and up) and work on web, desktop and mobile:
Claude Whiteboards (beta)
A shared board for brainstorming with your team and Claude. Talk through a messy plan and Claude turns it into a board with a timeline, owners and dates; in a retro it can group everyone’s notes by theme and draft action items. Drag a box around part of the board to ask about just that section. When you’re done, Claude can turn the board into a doc or deck, or send it to Lucid, Miro or Vani by Zoho.
Claude Sheets (beta)
Spreadsheets with tabs, formulas, formatting and drop-downs, often from one prompt (“build a hiring budget”). Claude outlines each cell it edits so you can track its changes, and you can share the sheet so teammates edit with Claude. Anthropic is upfront that it is simple spreadsheet functionality for now: for full Excel power, export to .xlsx or CSV, or use Claude for Excel.
Claude Data (formerly Claude Dashboards, now generally available)
Dashboards and “data stories” built on your company data through a data connector (a warehouse such as BigQuery or Redshift, or a CRM). New in beta: certified queries, so your data team can lock in the official definition of a metric (charts built on them carry a “certified” badge), and live sharing, so colleagues see current numbers without needing your connector access. Dashboards now cache and keep a version history you can roll back.
Two things to know before you plan around it. First, Claude Meetings — adding Claude to a video call to take notes and update your team’s documents while you talk — was announced the same day, but Anthropic’s own post calls it “coming soon.” Some news coverage described it as already launched; treat it as not yet available until it shows up in your account. Second, on Enterprise plans Whiteboards and Sheets start off (an admin turns them on in Organization settings > Artifacts); on Team plans all three are on by default.
Claude auto-invokes Artifacts for substantial outputs — code, long docs, HTML, SVG — and you can force one explicitly. It pops your work into a side panel next to the chat.
Predict first
Beyond looking nicer in a panel, what's the real win of an Artifact?
Do it · iterate in place
Refine an Artifact without re-pasting
Claude built you a landing page in an Artifact. Now just talk to it — each follow-up updates the same panel in place. Tap each refinement and watch it change.
landing-page.html⧉ ⤓
Your Product
The thing that does the thing.
0 of 4 refinements applied — tap one.
When to reach for it
Artifacts can also get in the way
The skill isn't forcing Artifacts on everything — it's knowing when they earn the panel.
Reach for an Artifact
Substantial work you'll iterate on or ship: code, long docs, an HTML mockup, an SVG, a React component. Force it with "make this an Artifact I can iterate on."
Skip it
One-paragraph answers, quick facts, a single calculation. An Artifact there is just friction.
The call
Which of these is worth forcing into an Artifact?
Ship it
Copy, download, deploy
Every Artifact has copy and download buttons. When you're done iterating, take the file straight out — HTML to a browser, code to your editor, Markdown to your CMS, SVG to your site. No re-formatting, no copy-paste-translate.
A classic code Artifact’s preview is sandboxed — it can't make network calls or touch your data. For real app testing, take the code out and run it locally. (The newer Claude Data type is the exception by design: it reads live company data through the data connector you give it.)
On very long Artifacts, Claude can lose track of state. If iteration starts breaking down past ~10–15 turns, paste the current Artifact back in fresh and keep going.
🛠️
Lesson complete
Claude is a workshop, not just a chat box
What you can do now
Force an Artifact when Claude doesn't auto-invoke one
Iterate across messages without re-pasting — refine in place
Tell when an Artifact helps vs when it just adds friction
Reset a runaway Artifact by pasting the current version back fresh
Ship Artifacts straight to file, code, or blog
Your move: build something small but real
Use Artifacts to build one small useful thing in a single session — a personal landing page, a recipe-saving script, an SVG icon set, a Markdown template. Iterate on it five or more times. Then copy it out and ship it.
Claude's long-context superpower: 200 pages, one prompt
Loading entire documents, comparing several at once, synthesizing research — the cases where Claude's context window genuinely changes what's possible.
Hey! I'm your AI Coach for this lesson on Artifacts. Ask me how to force one, how to iterate without re-pasting, when an Artifact helps vs hurts, or how to ship what you build. What are you trying to make?
Free lesson coaching is limited to 3 questions. Upgrade to Pro for unlimited coaching on every lesson.