Claude artifacts — the complete guide
What they are, what you can build, how to create and share one, where they fall short, and how to turn an artifact into a link anyone can open and comment on.
What a Claude artifact is
When you ask Claude to build or write something self-contained, it doesn't just print code into the conversation — it renders the result in a dedicated panel next to the chat. That working preview is the artifact. You refine it by asking ("make the header bigger", "add a dark mode"), and the panel updates in place.
Artifacts were built for the make-and-iterate loop: see the result, react, refine. The friction starts the moment the thing is good and you want someone else to see it.
What you can build as an artifact
Artifacts cover most self-contained outputs:
- Web pages and small apps — a landing page, a calculator, an interactive prototype, rendered live.
- Documents — a one-pager, a PRD, a brief, formatted and readable.
- Data visualizations — charts and dashboards from data you paste or describe.
- Diagrams and SVGs — flowcharts, simple illustrations, marketing graphics.
The common thread: anything that's better seen than read as code. If the output is a long back-and-forth conversation, it stays in the chat; if it's a discrete deliverable, it becomes an artifact.
How to create an artifact
You usually don't do anything special — ask Claude to build or write something self-contained and it renders an artifact automatically. To make it explicit, name the format: "make this a single self-contained HTML page" or "render this as a document I can preview." Self-contained is the key word — an artifact has to stand on its own, so prefer one file with no external dependencies.
How to share a Claude artifact
This is the part everyone hits, because inside Claude the artifact is yours and live, but outside your options narrow fast. There are three common routes, with real tradeoffs:
| Way to share | What the recipient gets | The catch |
|---|---|---|
| Copy the code out | A wall of HTML to save and open | Most people can't, or won't, open a raw file |
| Screenshot it | A flat image | Loses the layout on other screens, the links, and any way to change it |
| Publish to a link | A URL that renders in any browser | The artifact needs to be hosted somewhere persistent |
Publishing to a link is the only route that scales past one technical recipient. Once it's a link, the next question is whether people can react to it — point at a specific spot and leave a note — without an account.
Artifacts vs projects
A distinction people search for: an artifact is a single generated thing — one app, one doc, one chart. A project is a Claude workspace that holds context, files, and custom instructions across many conversations. You create artifacts inside projects; they aren't alternatives to each other. Use a project to keep Claude grounded in your context; the artifacts are what it produces along the way.
Where artifacts fall short
Honestly, two gaps show up repeatedly:
- Sharing and review. An artifact is a preview, not a destination. There's no built-in public link with anchored comments, so feedback ends up in screenshots and chat threads — exactly the workflow the artifact was meant to replace.
- Permanence and versioning. The artifact is tied to the session. If you want a stable URL that stays current as you iterate, you have to publish it somewhere that keeps version history.
Both are about what happens after you make something — the step from "I built it" to "the team signed off on it."
Making an artifact permanent and commentable
drafty.im/canvas/… URL that renders on any device. Anyone can click the exact element and leave a comment — no account — and your agent reads the comments and ships a new version on the same link, with version history. It works on artifacts from any tool, not just Claude.That's the gap between an artifact you made and a decision your team can sign off on: a link that renders anywhere, a way to point at the exact spot, and a loop back to the agent. However you choose to publish, that's the workflow to aim for.
Claude artifacts FAQ
- Are Claude artifacts public?
- No — by default an artifact is private to your session. To let others see one, you publish it to a link (e.g. to Drafty), which is where you choose who can view it.
- Can I share a Claude artifact without the other person having Claude?
- Yes, but not from inside Claude directly — you publish the artifact to a URL that renders in any browser, so the recipient needs nothing installed and no Claude account.
- What's the difference between an artifact and a project?
- An artifact is one generated thing (an app, a doc, a chart). A project is a workspace that holds context across many chats. You build artifacts inside projects.
- Do Claude artifacts have version history?
- Within the session you can iterate, but for a stable URL that keeps versions as you change it, publish the artifact to a tool that tracks versions.
- Can people comment on a Claude artifact?
- Not inside Claude. To collect anchored comments you publish the artifact to a link that supports element-level commenting, so reviewers point at the exact spot.