Blog / Guides
Guide

How to share Claude artifacts as a real link

Claude's built-in publish is great for a quick "look at this." The moment an artifact becomes a deliverable — a proposal, a report, a dashboard — you need a permanent URL, version history, and control over who sees it.

Claude can produce remarkably finished things inside a conversation: interactive dashboards, slide-style reports, calculators, single-page sites. Anthropic calls them artifacts, and if you've iterated on one until it's exactly right, you've probably hit the next question: how do I get this in front of someone who wasn't in the chat?

This guide covers every way to share a Claude artifact in 2026, where each one breaks down, and the workflow professional teams use when the artifact is going to a client, a stakeholder, or anyone whose opinion of you matters.

The four ways to share an artifact

MethodGood forWhere it breaks
Screenshot or screen shareA quick reaction in SlackStatic, no interactivity, dies immediately. Charts become pictures of charts.
Claude's built-in publishDemos, internal show-and-tellSnapshot per publish, no version history, no access control, and the URL announces "made with AI in one chat."
Copy the code to a static hostDevelopers with a deploy pipelineRepo, build step, DNS — a website project for what should be a share button. Non-technical teammates are locked out.
A dedicated sharing layer (Bauta)Client work, recurring reports, anything with your name on itRequires a one-time connector setup (about two minutes).

The first two are fine when the stakes are low. The rest of this guide is about what to do when they're not.

Why the built-in publish isn't enough for client work

To be clear: Claude's publish feature is well made, and for ad-hoc sharing it's exactly right. But it was designed for showing, not for delivering. Three gaps show up the moment an artifact becomes part of your work product.

1. Every update is a new link

Publish an artifact, tweak one number, publish again — you now have two URLs, and the one your client bookmarked is the stale one. You're back to the attachment problem in link form: re-sending, "use this one instead," and quietly hoping nobody opens the old version in a meeting.

2. No control over who opens it

A published artifact is open to anyone with the URL. For a portfolio piece, fine. For a proposal with pricing, a board pack, or client data, "anyone with the link" is a sentence your client's security team does not want to read.

3. It doesn't look like yours

The link lives on a claude.ai domain, framed as chat output. That's honest — but a deliverable should carry your identity, not your tooling's. Nobody sends clients a proposal from docs.google.com/preview and calls it a brand experience.

The professional workflow: publish from the conversation

The fix is not to leave Claude — it's to give Claude somewhere real to publish. Bauta connects to Claude as a connector (via MCP, the open protocol). After a one-time setup, the workflow is a single sentence:

You: "Share this." Claude: Published — your link: https://you.bauta.app/q3-review Private · you choose who can open it

Behind that sentence, a few important things happen:

Works beyond claude.ai. The same connector works from Claude Code, Cowork, ChatGPT, and Codex — anywhere your AI produces HTML. Set it up once and "share this" works across your whole toolkit. Setup guides: Claude Code, Cowork, ChatGPT.

A concrete example: the weekly client report

Say you produce a weekly performance report for a client as a Claude artifact — charts, commentary, a summary table. The old loop: export, rename report-week-26-FINAL.html, attach, apologize for the file size, repeat.

The new loop: the first week, you say "share this" and send the client one URL. Every week after, you refine the artifact and say "update the report." The client's bookmark always opens the current week; the version history quietly records every edition you shipped. You've replaced a recurring chore with a sentence.

And there's a quiet second benefit: because the report lives at a stable URL as structured HTML, the client's own AI tools can read it too. A link that both humans and agents can consume is worth more than a PDF in an inbox — more on that in why we think HTML beats PDF for AI-made documents.

When the built-in publish is still the right call

Honesty matters more than a pitch: if you're sharing a fun experiment with a friend, demoing something internally, or posting a toy to social media, Claude's built-in publish is the fastest path and you should use it. The dividing line is simple — will anyone open this link twice, and does it matter who? If yes to either, publish it somewhere permanent and gated.

Frequently asked questions

Can I share a Claude artifact with someone who doesn't have a Claude account?
Yes. Once published to Bauta, the artifact is a normal web page. Recipients open it in any browser — no Claude account, no login, unless you deliberately gate the link to specific people.
What happens to my shared link when I update the artifact?
The link stays the same. Claude republishes the new version to the same URL, and everyone sees the latest. Every previous version remains in history, so you can review or reference exactly what was live at any point.
Is it safe to share AI-generated HTML with clients?
On Bauta, shared content is served from an isolated sandbox domain (never our app domain), scanned against Google Safe Browsing, and hosted in the EU. Links are private by default. Read the full security architecture.
Do I need to know how to code?
No. The setup is pasting one connector address into Claude's settings — about two minutes, once. After that, publishing is conversational: "share this," "update it," "make it private."

Your next artifact deserves a real address

Connect Bauta to Claude once, then say "share this." Free plan, permanent links, no credit card.

Try it now