Capsuleby BrinkByte
Use case · Claude

Review Claude artifacts with your team

Claude builds an artifact in one conversation — a dashboard, a landing page, an interactive report. Then it lives and dies inside that chat. Capsule publishes Claude artifacts to a live share link where teammates, stakeholders and clients can review them: click around, comment on the page, and see each iteration land on the same link.

Claude artifacts are easy to make and hard to review

An artifact rendered inside Claude is visible to exactly one person: you. Getting feedback means screenshots, copy-pasting HTML into files, or asking five people to read a chat transcript. The feedback comes back as prose in Slack — detached from the element it describes — and you re-type it into the next prompt by hand.

That round-trip is the slowest part of an otherwise instant workflow. The generation took thirty seconds; the review took three days.

How the loop works

01

Generate

Build an artifact in Claude, or generate HTML with Claude Code.

02

Share

Publish to one private browser link — no zip files, no hosting setup.

03

Review

Reviewers click and comment directly on the live page.

04

Iterate

Your AI agent reads the feedback and proposes fixes; you decide what ships.

05

Publish

Approve the version and the same link updates for everyone.

Publish straight from Claude Code

Capsule connects to Claude Code (and any MCP-capable host) as an MCP server. One instruction — "publish this to Capsule" — and the artifact is live on a private share link. No export, no hosting, no zip files. Setup is a single command from the install page.

Reviewers comment on the live page

Anyone with the link sees the working artifact in their browser and leaves comments anchored to the part of the page they mean. Password-protect or time-limit the link if the audience is external.

Claude reads the feedback and proposes fixes

Back in Claude Code, your agent collects the open comments, suggests a fix for each, and waits for your call. Approve, and the new version replaces the content behind the same link — reviewers see what changed, and each addressed comment is queued for them to accept.

Frequently asked questions

How do I share a Claude artifact with someone outside the chat?

Publish it to Capsule from Claude Code over MCP (or paste the HTML into the REST API). You get a live browser link immediately — optionally password-protected — that anyone can open without a Claude account.

Can Claude act on the review feedback automatically?

Claude reads the comments and proposes fixes, but Capsule is built around human approval: you decide which fixes ship. Nothing is published without the builder approving it.

Does this work with Claude Code or only claude.ai artifacts?

Both. Claude Code connects over MCP and can publish, update and read comments directly. HTML from claude.ai artifacts can be published the same way, or via the REST API.

What happens when I iterate on the artifact?

Updates are pushed as new versions of the same capsule. The share link never changes, version history is kept, and reviewers are never looking at a stale copy.

More review workflows: Review Cursor projects · Review Lovable apps · Collaborative review for AI-generated apps

Collaborate around AI-generated apps — not screenshots

One live link for the page, the feedback and every version.