How to share a Claude artifact

The one-click path: Claude's built-in Publish

Inside Claude, every artifact has a Publish button at the bottom of the panel. Click it and Claude generates a public URL. The artifact renders in any browser; the chat that produced it stays private.

What the native publish does well:

• No account required to view. Recipients open the link, see the artifact, and can interact with it. They're prompted to sign up only if they want to remix it (fork a copy into their own Claude).
• Instant. No export step, no file to upload.
• Free on all plans — Free, Pro, and Max can all publish publicly.

What it doesn't do:

• No expiry or password on the link
• No view analytics
• Published links may be indexed by search engines
• Once you unpublish, you can't republish the same artifact — Anthropic permanently deletes the associated storage. If you need to revise and reshare, you'll create a new URL.
• Team and Enterprise users cannot publish publicly. Sharing on those plans is organization-only; anyone viewing must be authenticated to your org.

If the publish button covers your case — a quick share to a colleague who's fine with a permanent public link — you're done. The rest of this guide covers the scenarios where it doesn't.

When the publish link isn't enough

Three situations keep coming up:

You need to send a client something that doesn't live on Claude's domain forever. A public artifact at claude.ai/artifacts/… can be indexed and seen by anyone who finds the URL. For a proposal or a prototype you'll iterate on, that's a problem.

You need to update the artifact without changing the URL. Claude's publish doesn't support in-place updates. Revise the artifact, and the only option is a new link — now the client has two URLs and no clear signal of which is current.

You need feedback that isn't "the button feels off." Reviewers leave comments in Slack, email, a Notion doc. None of those comments are attached to the specific element they're about. You're left guessing whether "the header" means the navigation or the hero text.

All three point at the same gap: the artifact is easy to generate, but the review step that makes it useful to others is still held together with screenshots.

Three routes for sharing an artifact

Route 1: Native Publish

Covered above. One click, zero setup, no expiry control.

Route 2: Download and host elsewhere

Claude lets you download the artifact as an .html file. From there, you can upload it to any static host — or to services like ShareDuo that add expiry windows (1 hour to 30 days), optional passwords, and view analytics.

The tradeoff: the artifact is now a static file. If the original used Claude's API at runtime, that functionality won't survive the export. And you're back to managing files and links manually.

Route 3: Publish to a canvas and get feedback

The difference between routes 2 and 3 is the feedback loop. A static host or the native publish link gets you a URL; a canvas gets you a URL that reviewers can annotate at the element level, with comments routed back to Claude so the next iteration happens in the tool that made it.

This matters most when the artifact is a deliverable — a prototype a client needs to sign off on, a report a PM needs to approve — rather than a quick demo.

A realistic workflow for makers and PMs

Here's what the loop looks like when you're iterating with a client or team:

1. Build the artifact in Claude. A landing page concept, a dashboard prototype, a one-pager.
2. Push it to a canvas link. One command if you're using the Drafty CLI (drafty push artifact.html), or paste the HTML.
3. Send the link. The client opens it in their browser — no Claude account, no signup.
4. Client clicks the exact element and leaves a note. "Make this section shorter." "The CTA color needs to match the brand."
5. Your agent picks up the comments and revises. The updated version lands on the same URL. The client refreshes and sees the change.

The thing that breaks down in the screenshot-and-Slack version of this workflow is step 4 — "the section" could mean three different things, and you find out which one only after guessing wrong once. Pinning the comment to the element removes that round-trip.

What "no Claude account needed" actually means

It means the recipient doesn't need to pay for Claude or log into claude.ai to see your artifact. It doesn't mean the artifact runs forever with zero dependencies.

A few things still matter:

• AI-powered artifacts require sign-in. If your artifact calls Claude's API at runtime (for dynamic responses, real-time generation), viewers must sign in to Claude to use those features. Static HTML artifacts — pages, charts, prototypes built as self-contained files — work for anyone.
• The link is only as permanent as where you host it. Claude's native published links are permanent until you unpublish; once unpublished, the URL is gone for good. A canvas link is as permanent as the canvas.
• Team/Enterprise users need a different path. If you're on a Claude Team or Enterprise plan, the native Publish option doesn't produce a public link. You'll need to export and host elsewhere, or use a tool that isn't gated behind Claude's org auth.

The one thing most people get wrong

They share the published link and then iterate the artifact in Claude — generating v2, v3, v4 — while the client is looking at the original published URL, which hasn't changed. There's no signal to the client that a new version exists, and there's no way to update the link in place.

The fix is to decouple the "link you share" from "the artifact inside Claude." Publish to somewhere that lets you push updates to the same URL. Then the client always has the right version.