Claude Artifacts: Twelve Examples Worth Building

The Claude artifacts worth building are the ones you would otherwise make by hand and then keep updating: a dashboard from data you already have, a calculator for a sum you do every week, a checklist that remembers what you ticked, a document or deck to send, a small tool, a diagram. Anthropic’s help centre defines an artifact↗ (opens in a new tab) as anything Claude makes that you would put in front of someone, opened beside the conversation so you can edit it, come back to it and share it. Below are twelve, each with a one-line prompt and what you get. Who can open one once it is shared, and where that stops being enough for a team, is covered in Claude Artifacts sharing limits.

Dashboards

• Project status. Prompt: “Here is our board as Markdown. Build a one-page status dashboard: counts per lane, bugs open, and the five oldest items in progress.” You get a page of counts and lists you can reload with fresh text next week.
• Metrics from a file. Prompt: “From this CSV of weekly sign-ups, build a dashboard with a trend line, week-on-week change and a table I can sort.” You get charts and a sortable table, drawn from the file you attached rather than typed-in numbers.
• A pull-request walkthrough, in Claude Code. Prompt: “Make an artifact that walks through this PR with the diff annotated inline.” Anthropic’s Claude Code artifacts documentation↗ (opens in a new tab) gives this as a standard use: a page at a private claude.ai address that updates in place as the session continues.

Calculators

• Quote calculator. Prompt: “Build a quote calculator: day rate, days, expenses and a discount slider, with a total and a line I can copy into an email.” You get inputs, a live total and a copy button.
• Meeting cost. Prompt: “Make a calculator for what a meeting costs: attendees, average salary, length, and how often it repeats.” Small, but the kind of thing people open again, which is the test of a good artifact.

Checklists and trackers

• A personal launch checklist. Prompt: “Build a launch checklist for a small web app with items I can tick, and keep my ticks between visits.” On paid plans an artifact can store data, and personal storage keeps each person’s ticks private to them.
• A shared rota. Prompt: “Make an on-call rota where each person can claim a week, and everyone sees the same rota.” That uses shared storage: everyone sees the same data, and the first time someone uses it Claude shows a confirmation that their data will be visible to others.

Documents and decks

• An onboarding doc. Prompt: “Start a doc: onboarding for a new support hire, first week day by day.” With the Docs template you get a living document you and your team write in with Claude, exportable to Word, PDF, Markdown or Google Docs.
• A deck from notes. Prompt: “Turn these meeting notes into a six-slide update for the leadership team.” With the Slides template you get slides you can edit directly, present without leaving Claude, and export to PowerPoint or PDF.

Small tools and diagrams

• A quiz that uses Claude. Prompt: “Build a flashcard quiz on our product’s pricing rules that explains each wrong answer.” Artifacts can call Claude directly; people who use it sign in with their own Claude account, and their use counts against their own plan limits rather than yours.
• A meeting-prep sheet from your calendar. Prompt: “Build a page that lists tomorrow’s meetings from my calendar with a notes box for each.” On paid plans, artifacts can connect to apps you have connected to Claude, such as Google Calendar, after you approve which tools they may use.
• A process diagram. Prompt: “Draw our expense approval process as a flowchart: submit, manager check, finance check, paid, with the rejection paths.” You get a diagram you can correct by describing the change, which is quicker than moving boxes by hand.

What artifacts can do today

As documented on 28 September 2026, and marked beta where Anthropic says so:

• Plans. Artifacts are on Free, Pro, Max, Team and Enterprise. Templates (Docs, Slides and Design) are in beta on paid plans only. Artifacts need code execution and file creation turned on in Settings, Capabilities, or in organisation settings on Team and Enterprise.
• Run as small apps. An artifact can be a single-page website or an interactive tool, and it can call Claude, so the page answers questions or generates content for whoever is using it.
• Remember things. On Pro, Max, Team and Enterprise, on the web and desktop, artifacts can store data between sessions, personal or shared. The limit is 20 MB per artifact, text only.
• Use your apps. On the same plans, artifacts can read from and write to connected apps and custom connectors. Everyone connects their own apps, even on a shared artifact, and a connector tool that needs approval for every action is not available to artifacts.
• Be found again. Everything you make is kept in the Artifacts tab in the sidebar, and on mobile you can view artifacts there, though starting from a template, editing and sharing settings need the web or desktop app.
• Come from Claude Code. Claude Code can publish session output as an artifact on Pro, Max, Team and Enterprise, and each publish becomes a version you can choose to show.

What they cannot do

• Run a backend. A Claude Code artifact is one self-contained page with no server; it cannot serve several routes or authenticate viewers itself. For a hosted internal tool with a backend, deploy it on your own infrastructure.
• Load anything from anywhere. Claude Code artifacts load scripts only from a short list of public CDNs, fonts only from Google Fonts, and embed images inline.
• Store files. Storage is text only: no images, files or binary data.
• Share everything by link. Anthropic’s page on sharing artifacts↗ (opens in a new tab) says artifacts that connect to your apps or use Claude cannot use “Anyone with the link”; on Team and Enterprise plans, docs cannot yet either.
• Behave the same if they are old. Artifacts made in a chat before 16 September 2026 are legacy artifacts: they keep working and can still be published or shared, but their storage works only once published, and no new ones can be made.

How to iterate without starting again

• Say what the data looks like before asking for the page. “Columns: date, plan, country” gets a better first draft than “my data”.
• Change one thing per message. “Add a per-region breakdown under the chart” is easy to check; five changes at once are not.
• Point instead of describing. In a doc or Markdown document, highlight the passage, choose “Edit with Claude” and type the change, so Claude edits exactly there.
• Branch rather than overwrite. Editing an earlier message creates a new version of the chat with its own artifacts, so you can try another direction and keep the first.
• Let errors come back to Claude. If an artifact shows an error, “Try fixing with Claude” copies the details into a new message.
• Plan for the viewer. For a page that reads connected apps, ask for a fallback message naming the connector each section needs, since each viewer uses their own connections. For a shared tracker, check whether it uses shared storage before anyone types something private into it.
• Only open artifacts from people you trust. Anthropic’s help centre says to treat someone else’s artifact the way you would treat a file from an unknown sender.

Where the work list should live

The first example works best with the list kept somewhere else. On a fenbs board, Copy as Markdown puts every lane (To Do, Next Up, In Progress, Completed) on your clipboard as a Markdown list; paste it into Claude with the prompt above and the artifact becomes a picture of the board, rebuilt whenever you paste again. The board stays the record: each task is a feature, an enhancement or a bug, people and AI assistants are members with roles, and every change carries the name of whoever made it. An artifact’s shared storage records none of that. If you have added fenbs to Claude as a connector, a dashboard artifact could read the board through each viewer’s own connection, which also means it can show a viewer only what their role on the board allows.