# What is a Claude artifact? · Stacktree Source: https://stacktr.ee/what-is-a-claude-artifact [Skip to content](#main) [Stacktree](/)[Developers](/developers)[Agents](/agents)[Docs](/docs)[Use cases](/use-cases)[Pricing](/pricing)[Blog](/blog)[Dashboard](https://app.stacktr.ee)[Sign in →](https://app.stacktr.ee) By [ Steve Smith ](/about) · Founder, Stacktree · Last updated September 9, 2026 Claude artifacts, explained # What is a Claude artifact? It is the panel that opens beside the chat when Claude builds something substantial: a document, a web page, a chart, a small working app. It updates in place as you refine it, which is what makes it feel like a thing rather than a message. Here is what artifacts do well, where they actually live, and the four limits that decide whether you can send one to anybody. [ Get started free ](https://app.stacktr.ee/?join=1&from=seo_what_is_a_claude_artifact) install npx stacktree-install Copy No card · 3 pages free · about a minute ## What is a Claude artifact? A Claude artifact is a self-contained piece of work — a document, a web page, a diagram, a small app — that Claude builds in a panel next to the conversation instead of writing into the chat. It updates in place as you ask for changes, so you refine one thing rather than scrolling through versions of it. It lives on Anthropic’s servers, attached to your account and the conversation that made it, and it is private until you publish it. ## On this page - 01 [ What Claude puts in an artifact ](#what) - 02 [ Why it is a panel and not a message ](#why-panel) - 03 [ Where an artifact actually lives ](#where) - 04 [ Who can open one ](#who) - 05 [ The four limits worth knowing ](#limits) - 06 [ When an artifact should become a page ](#when) ## What Claude puts in an artifact Claude reaches for an artifact when the output is substantial, standalone, and something you are likely to edit or reuse rather than read once. In practice that means: - Documents. Reports, specifications, long-form writing, anything with structure you will come back to. - Web pages. A single file of HTML, CSS and JavaScript that renders live in the panel. - React components. Interactive interfaces that actually run rather than being described. - Diagrams and charts. SVG, flowcharts, sequence diagrams, data visualisations. - Small tools. Calculators, dashboards, explainers, games. Anything self-contained enough to work without a server behind it. The unifying idea is one self-contained file. An artifact is something that reads or runs on its own, with no build step and nothing to install. That constraint is why artifacts appear so fast, and it is also the source of most of their limits. ## Why it is a panel and not a message The panel is not a display decision, it is a versioning one. Chat is append-only: ask for a change and you get a new message, leaving you to work out which of the six blocks of code on screen is the current one. An artifact replaces itself. Ask for a darker header and the header goes dark in the thing you are already looking at. That single behaviour is what makes artifacts feel like work rather than conversation, and it sets an expectation that matters later. Once you have felt a thing update in place, going back to emailing `report-v3-FINAL.html` to somebody is genuinely painful. ## Where an artifact actually lives This is the part most explanations skip, and everything practical follows from it. An artifact lives on Anthropic’s servers, attached to your account and to the conversation that produced it. It is not a file on your computer. It is not a website you own. It is a view into Claude, and reaching it normally means reaching your Claude account. For the thing artifacts are designed for — building something with Claude and looking at it — that is exactly right, and the speed comes from it. The friction starts when the artifact stops being something you are looking at and becomes something you need to hand over. ## Who can open one By default, only you, inside the conversation. Publishing changes that: Claude mints a claude.ai link and anyone holding it can open the artifact in an ordinary browser, with no Claude account needed. What is missing is the middle. There is no per-person access, no passcode, and no way to restrict a link to one company’s email domain. The choice is private to you, or open to anyone who ends up with the URL — including anyone your recipient forwards it to. We have written up [whether published artifacts are public](/make-claude-artifact-private) and what to do when that is not the access model you wanted. ## The four limits worth knowing None of these are complaints about artifacts. They are consequences of the design, and they only bite in one situation: when the artifact is a deliverable rather than a demo. - The address is not yours. A published artifact sits on claude.ai. For internal work that is fine. When the recipient is a client, the domain is part of what you are delivering. - Access is all or nothing. Private to you, or open to anyone with the link. No passcode, no named recipients, no company-domain restriction. - Unpublishing is one-way. Per Anthropic’s help centre, once you unpublish, the Publish button deactivates and that version cannot be restored; if the artifact used persistent storage, that data is permanently deleted. We cover [the unpublished-and-cannot-republish message](/blog/claude-artifact-cannot-republish) separately. - It depends on somebody else’s uptime. When you see artifact storage is temporarily unavailable, nothing you did caused it and waiting is usually the fix. But if that artifact was the thing due this afternoon, the outage is still your problem. ## When an artifact should become a page The honest rule: leave it as an artifact while you are building and showing. Move it when you are delivering. Four signals that the moment has arrived. It needs to sit on your own domain. It must reach one named person and nobody else. It has to outlive the conversation and the account that made it. Or it will be revised, and whoever you sent it to must keep seeing the current version at the address you already gave them. Plain disclosure, since this is our own site: Stacktree does that last part. You publish the artifact’s HTML and get a private, unguessable URL that opens with no account for the reader, can carry a passcode or a company-email gate, can sit on your own domain, and can be replaced in place so a link you sent last week shows this week’s version. An agent can do it directly through [MCP](/mcp-publish-html), or you can move one by hand in about thirty seconds — the walkthrough is at [how to share a Claude artifact](/share-claude-artifacts). If none of those four signals apply to what you are making, publish it from Claude and get on with your day. That is what the button is for. FAQ ## Claude artifacts: common questions What is a Claude artifact? + An artifact is a self-contained piece of work Claude builds in a panel beside the conversation instead of writing into the chat: a document, a web page, a diagram, a small app, a chart. Claude creates one when the output is substantial, standalone and something you are likely to edit or reuse rather than read once. The practical difference from ordinary chat output is that an artifact updates in place as you refine it, so you watch one thing improve rather than scrolling through six versions of it. What can a Claude artifact contain? + Documents and long-form writing, single-file web pages with HTML, CSS and JavaScript, React components, SVG diagrams and charts, and small interactive tools such as calculators, dashboards and games. The useful mental model is one self-contained file: an artifact is a thing that runs or reads on its own, without a build step or a server behind it. Where does a Claude artifact actually live? + On Anthropic's servers, attached to your account and to the conversation that produced it. That is the single most important thing to understand about artifacts, because everything else follows from it. The artifact is not a file on your computer and not a website you own. It is a view into Claude, and reaching it normally means reaching your Claude account. Are Claude artifacts public? + Not by default. A new artifact is visible only to you inside the conversation. Publishing it creates a claude.ai link that anyone holding the link can open, which is a deliberate act and not the default state. What you cannot do is publish to a middle ground: there is no per-person access, no passcode, and no restriction to one company's email domain. We cover the detail on whether published artifacts are public. Can someone open a Claude artifact without a Claude account? + Yes, once it is published. A published artifact link opens in an ordinary browser for whoever holds it. Before publishing, no: the artifact exists inside your account, and the way people usually attempt to share one at that stage is a screenshot, which loses everything interactive about it. What happens when I click publish on an artifact? + Claude mints a shareable claude.ai link to that version and makes it readable by anyone with the link. It is genuinely quick, and for showing a colleague something you just made it is the right tool. It is not website hosting: the page stays on Anthropic's domain, it stays tied to your account and the conversation behind it, and you cannot put it on a domain of your own or brand it. Can I unpublish a Claude artifact and publish it again? + You can unpublish, but per Anthropic's help centre the action is one-way for that version: the Publish button deactivates and that published version cannot be restored. If the artifact used persistent storage, unpublishing permanently deletes the associated storage data. Worth knowing before you unpublish something you have already sent to a client. What does "artifact storage is temporarily unavailable" mean? + It is a service-side error, not something you have broken, and the usual first move is simply to wait and retry rather than rebuild anything. It surfaces most often for artifacts using persistent storage. The reason people search it in a panic is the same reason it is worth knowing about: when the artifact is the deliverable and it is due, an outage on somebody else's infrastructure is your problem. A static copy on an address you control is not affected by it. How is an artifact different from a Claude Code project? + An artifact is one self-contained thing produced inside a chat. Claude Code works against a real filesystem, so it produces directories, dependencies and builds. The two overlap when Claude Code writes a single HTML page, but a Claude Code output is generally a project you own on disk, while an artifact is a panel inside a conversation. When should I move an artifact somewhere else? + When it stops being a thing you are showing and becomes a thing you are delivering. Four signals: it needs to sit on your own domain rather than claude.ai; it must be private to one recipient rather than open to anyone with the link; it has to survive independently of your Claude account and the conversation it came from; or it will be revised and the person you sent it to must keep seeing the current version at the same address. Keep reading ## Related guides - [ How to share a Claude artifact The thirty-second version: artifact to private link, with a passcode if it needs one. ](/share-claude-artifacts) - [ Are published artifacts public? What publishing actually exposes, and how to get an access model with a middle setting. ](/make-claude-artifact-private) - [ Unpublished and cannot be republished Why that message is final, and what to do with the artifact now. ](/blog/claude-artifact-cannot-republish) - [ Public sharing and multiplayer editing What Anthropic shipped in July 2026, and the three things it still does not do. ](/blog/claude-artifacts-public-sharing) - [ Live artifacts explained Artifacts that fetch and hold state, and where the boundary sits. ](/blog/what-are-claude-live-artifacts) - [ Artifacts vs Claude Code When the output is a project on disk rather than a panel in a chat. ](/claude-code) References ## Sources and further reading - [ Anthropic: publish and share artifacts ↗ The help-centre article documenting what publishing does, and that unpublishing is one-way for that version. ](https://support.claude.com/en/articles/9547008-publish-and-share-artifacts) - [ Anthropic: create and use artifacts ↗ Anthropic’s own definition of what an artifact is and when Claude produces one. ](https://support.claude.com/en/articles/9487310-what-are-artifacts-and-how-do-i-use-them) ## When the artifact becomes the deliverable. Publish it to a private link on your own domain, gated if it needs to be, replaced in place when it changes. Free to try, no account for whoever opens it. [ Sign up free → ](https://app.stacktr.ee/?join=1&from=cta_what_is_a_claude_artifact) install npx stacktree-install Copy Private hosting for the HTML your agents make. [](https://betalist.com/startups/stacktree?utm_campaign=badge-stacktree&utm_medium=badge&utm_source=badge-featured)[Featured on](https://devhunt.org/tool/stacktree)[](https://buildlist.io)[Get started free](https://app.stacktr.ee/?join=1) ## Product - [Example deliverables](/examples) - [Custom domains](/custom-domains) - [Client feedback](/client-feedback-loop) - [One-time links](/one-time-view-links) - [See how pages get read](/page-engagement) - [Made with Stacktree](/made-with) - [Security](/security) - [Watch the demo](/demo) ## Agents - [All integrations](/agents) - [Claude Code](/claude-code) - [OpenAI Codex](/codex) - [Cursor](/cursor) - [Claude.ai connector](/claude-ai-connector) - [MCP server](/mcp-publish-html) - [Deploy from Claude Code](/deploy-html-from-claude-code) - [Skills](/skills) - [Slack app](/slack) - [n8n node](/n8n) - [Agent payments (x402)](/x402) ## Alternatives - [All comparisons](/alternatives) - [Head-to-head comparisons](/compare) - [Tiiny Host](/tiiny-host-alternative) - [GitHub Pages (private)](/github-pages-private-alternative) - [Vercel](/vercel-alternative-for-agents) - [ngrok](/ngrok-alternative-for-html) - [Display.dev](/display-dev-alternative) - [Static.app](/static-app-alternative) - [OpenAI Codex Sites](/openai-codex-sites-alternative) - [here.now](/here-now-alternative) - [Shippage](/shippage-ai-alternative) - [Best private hosting](/best-private-html-hosting) ## Use cases - [All use cases](/use-cases) - [Share with clients](/share-with-clients) - [Send a file to a client](/share-html-file-with-client) - [Share Claude artifacts](/share-claude-artifacts) - [Share Jupyter notebooks](/share-jupyter-notebook-html) - [Host Storybook privately](/host-storybook-privately) - [Architecture diagrams](/share-architecture-diagrams) - [AI-generated reports](/host-ai-reports) - [Internal HTML tools](/internal-tool-hosting) - [Private HTML hosting](/private-html-hosting) - [Vibe-coded page hosting](/vibe-coding-hosting) - [Leave a website builder](/website-builder-migration) ## Learn - [Blog](/blog) - [Glossary](/glossary) - [FAQ](/faq) - [Agent-loop hosting](/agent-loop-hosting) - [Why agents need a publish primitive](/blog/why-agents-need-a-publish-primitive) - [MCP servers explained](/blog/mcp-servers-explained-for-developers) - [What changed in the 2026-07 MCP spec](/blog/mcp-2026-spec-changes) - [Sites in Codex explained](/blog/sites-in-codex-explained) - [Private-by-default hosting](/blog/private-by-default-html-hosting) - [An agent paid us $1 (x402)](/blog/agent-paid-to-provision-itself) - [When a loop hits a paywall](/blog/loop-engineering-paywall) - [Pricing](/pricing) - [Self-host (new)](/self-host) - [Changelog](/changelog) - [Docs](https://stacktr.ee/docs) - [About](/about) © 2026 Stacktree · stacktr.ee [Privacy](/privacy)[Terms](/terms)[Security](/security)[Dashboard](https://app.stacktr.ee)[npm](https://www.npmjs.com/package/stacktree-mcp)[Sitemap](/sitemap.xml)[llms.txt](/llms.txt)[API spec](/openapi.json) --- Full markdown summary of the Stacktree marketing surface: https://stacktr.ee/llms-full.txt