# How to Share a Claude Artifact and Actually Get Feedback on It

Claude now has a Publish button. Click it, get a public link, send it to anyone — they open it in a browser, no Claude account needed. It works, and for a lot of use cases, it's the whole answer.

This post is about the use case where it isn't.

## What Claude's Publish does well

Credit where it's due: Publish solves the basic sharing problem cleanly. You built something — a calculator, a dashboard, a landing page, a little game — and you want someone to see it. Click Publish, copy the link, send it. Done.

Recipients can view it, fork it into their own Claude conversation, and you can embed it on another site. If you just want to show someone a thing you made, this is the right tool and you should use it.

## Where it runs out of road

Publishing is broadcasting. You put it out there, and then... you wait. Maybe someone texts you "looks good." Maybe someone says "the button thing is weird" and you spend ten minutes figuring out which button, on which screen, in which browser.

Three things Publish doesn't do:

**1\. You don't know if anyone looked at it.** No view count, no analytics, no read receipts. You send the link and hope.

**2\. There's no way to leave feedback on the page itself.** The person looking at your artifact has to switch to a different channel — Slack, text, email — and describe what they're reacting to in words. "The thing on the left, no the other left, the blue one, wait is yours blue? Mine's grey." Every prototype review in history has had this conversation. Publishing the artifact doesn't fix it.

**3\. Feedback doesn't connect back to Claude.** Someone tells you "make the hero bigger." You open Claude, retype what they said from memory, and hope you captured it right. The feedback lives in a text thread. The fix happens in Claude. There's a human copy-paste step in between, and that step loses context every time.

## The version where feedback is built into the link

This is the part where I talk about [dot.](https://leaveadot.com), because I built it for exactly this gap. Full disclosure: I'm the founder, and everything below is the product working as designed — not a hypothetical.

The flow:

**You build something in Claude.** A landing page, a prototype, a dashboard, whatever. Then you say "share this with dot." The MCP connector gives you a live, hosted link — same as Publish, except the link has a feedback layer on it.

**You send the link to anyone.** They open it in a browser. No Claude account, no dot. account, no extension, no signup. They see the live artifact exactly as Publish would show it.

**They click anywhere on the page to leave a comment.** Directly on the element they're reacting to. Not in Slack, not in a text thread, not in a Google Doc with arrows drawn in red. On the thing itself. A pin drops where they clicked, and their comment is anchored to that spot.

**Every comment carries context.** A snapshot of exactly what they were looking at — the button, the card, the section — plus their browser, OS, and screen size. So "this looks broken on mobile" arrives with a screenshot of the broken thing and the device it was broken on.

**Then the part nothing else does.** Back in Claude, you say "pull my dot. feedback and fix what reviewers flagged." Claude reads every pinned comment — with the element snapshot and environment context — and regenerates the artifact with the fixes applied. The person who said "make the hero bigger" just talked directly to the AI that built it, without either of them knowing the other existed.

## What about Team and Enterprise plans?

Worth mentioning: Claude Team and Enterprise accounts can't publish artifacts publicly at all. The Publish button creates an internal-only link visible within your organization. If you need someone outside your org — a client, a stakeholder, a freelancer — to see the artifact, Publish doesn't work.

dot. doesn't have that restriction. The review link works for anyone, on any plan, inside or outside your organization.

## What about the other sharing tools?

There's a growing list of tools for hosting Claude artifacts — Stacktree, Shareable, ShareDuo, and others. They solve real problems: private hosting, password protection, analytics, custom domains. If you need those, they're good options.

What they have in common: they're all variations on "host the artifact somewhere and give people a link." The thing people do after they open the link — react to it, point at specific parts, tell you what to change — is outside their scope.

dot. is the only one where the link itself is commentable, and the only one where those comments flow back into Claude as structured input it can act on.

## The short version

Claude's Publish button: share a link, hope for the best.

dot.: share a link, people pin feedback on the live page, you pull it back into Claude and fix everything in one prompt.

Same artifact. Different outcome.

* * *

*Built something in Claude?* [*dot.*](https://leaveadot.com) *gives it a link where anyone can leave feedback — pinned on the page, with full context, no account needed. Then Claude fixes what they flagged.* [*Try it free →*](https://leaveadot.com)