Claude Artifacts: what they are, what you can build and how to share them

What are Claude Artifacts?

Anthropic's help center article on artifacts defines an artifact as anything Claude makes for you that you would put in front of someone: a design, a deck, a document, a dashboard or a small interactive tool. It opens beside your conversation, you can edit it and come back to it, and you can share it. You ask for one in any chat, or start from a template in the Artifacts tab in the sidebar.

The current help article separates new artifacts, which are built from templates on a canvas or page you can work in directly, from legacy artifacts, which are the ones made in a chat before September 16, 2026. Legacy artifacts keep working and can still be published and shared, but you cannot make new ones. Everything below describes the current kind and flags where legacy ones differ.

What can you make with Claude Artifacts?

Templates are starting points for work you will share. Ask for one in any chat, select Output in the message box and choose a template, or pick one in the Artifacts tab. The table follows the help article as of October 2026.

Beyond the templates, the article says Claude can make an artifact out of almost anything that stands on its own, including documents, code snippets, single-page websites, images, diagrams and flowcharts, dashboards and small interactive tools. That last group is the one builders care about, because a small tool is a prototype. A tip calculator, a quiz, a habit tracker or a unit converter is a few minutes of prompting.

How do you create and edit an artifact?

1. Open Claude on the web or in the desktop app and describe the thing you want in plain language. Say who it is for and what it should do. Or choose a template under Output.
2. Claude builds it and opens it in a window beside the chat. Artifacts made from a template open on a canvas or page.
3. Change it by asking Claude, or edit directly: type in a doc, edit a slide, move elements on a design canvas. In a doc or Markdown document you can highlight text, choose Edit with Claude, and say what to change. Claude edits where you marked it.
4. If the artifact shows an error, use the Try fixing with Claude button near the message. It copies the error details into a new message. Anthropic notes that success is not guaranteed.
5. To explore another direction without losing the first, edit an earlier message in the chat. That creates a different version of the conversation with its own artifacts.
6. Find old work later in the Artifacts tab in the sidebar, which lists everything you have made.

On the iOS and Android apps you can ask for a design, deck or doc in a chat, view it in the Artifacts tab and edit it. To change sharing settings, use Claude on the web or the desktop app.

Can an artifact call Claude, store data or connect to your apps?

Yes to all three, with plan limits. These are what turn an artifact from a static page into a small app.

Two cautions from the same article. Whoever builds the artifact decides which data uses personal storage and which uses shared, so look before you enter anything sensitive into someone else's tool. And on legacy artifacts, storage only works after you publish, and unpublishing permanently deletes the stored data.

How do you share a Claude artifact?

Artifacts start private to you. Open the artifact, click Share, choose who has access, copy the link and send it. The options depend on your plan, per the sharing article.

For people you invite, there are three access levels. Can view lets them open the artifact and read comments. Commenter adds comments and downloading any files the artifact offers. Can edit adds making changes. Anyone you share with needs a Claude account to open the artifact; the one exception is a legacy chat artifact that was published. You cannot change sharing settings on the iOS or Android apps. Documents made with Claude Docs, and artifacts that connect to other websites, cannot be shared by email invitation.

Because viewers need an account, an artifact is a poor way to put a product in front of the general public. It suits a teammate, a client, a friend or a small group. If you want something strangers can find, you need a real website.

What about artifacts in Claude Code?

Claude Code can publish session output as an artifact: a live page at a private URL on claude.ai that updates as your session continues. The Claude Code artifacts page gives examples such as a pull-request walkthrough with annotated diffs, a dashboard built from data the session pulled, and an investigation timeline that fills in while a long task runs.

It also states the limits plainly. An artifact there is one self-contained page with no backend, so it cannot serve multiple routes or authenticate viewers; for a hosted internal tool, deploy to your own infrastructure instead. The page runs under a strict content security policy, so it loads scripts only from a short list of public CDNs, inlines other CSS and JavaScript, and cannot start a download by itself. The rendered page must be 16 MiB or smaller. Publishing needs a Pro, Max, Ultra, Team or Enterprise plan, a session signed in with a claude.ai account (not an API key or a cloud-provider credential), and the Anthropic API as the model provider. It does not work on Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry. For Claude Code basics, see the Claude Code tutorial.

When is an artifact enough, and when do you need a real app?

An artifact is the fastest way to find out whether an idea feels right. It is not a product you can run a business on. Use this table to decide.

When an artifact proves the idea, the next step is an app you own. The guide to building an app with Claude Code covers that jump, and Claude Code vs Cursor vs Lovable compares the ways to do it. If you do not have an idea yet, the board has plenty to start from. For example, a small shared tool like Kids chore board with pocket money was posted on the board, and its shape (a few screens, shared data, no payments) is the kind of thing an artifact can prototype in an afternoon. Browse more under productivity ideas.

What prompts work well for a first artifact?

A good artifact prompt names the audience, the screens or sections, the data it keeps and what it should not do. Start from this and swap in your own subject.

Build an interactive artifact: a household chore tracker for two adults and two kids.

Screens:
1. Today: each person's chores with a checkbox.
2. Week: a grid of who did what.
3. Settings: add or remove people and chores.

Storage: shared, so everyone using the artifact sees the same chores. Text data only.

Rules: no sign-up screen, no payments, works on a phone-width screen. Keep the design plain and readable. When you finish, list the three things you would change first.

Create a Doc artifact: a one-page brief for a new teammate about how our weekly release works.

Sections: What ships and when, Who approves, What to do if a release breaks, Where to ask for help.

Use short sentences. Leave clearly marked placeholders where you need facts from me, and do not invent names or dates.

The second prompt carries a rule worth copying into every artifact request: tell Claude to leave placeholders instead of inventing facts. Claude can write a convincing but wrong date or name into a document, and nobody notices until it is shared.