Claude Artifacts adds public links for generated prototypes

• Claude Artifacts now has public links, multiplayer editing in Claude Code, and Claude Tag creation, according to ClaudeDevs' launch thread.
• Public publishing moves generated prototypes beyond the workspace because anyone with the link can view the artifact, ClaudeDevs' public-link note says.
• Team editing is the paid-team piece: ClaudeDevs' launch thread says multiplayer editing is available on Team and Enterprise plans.
• Claude Tag can turn a Slack request into a dashboard artifact, and trq212's dashboard workflow described a loop where others or local Claude Code sessions edit that dashboard.
• A creator-side edge case remains React versus HTML: petergyang's reply said one edit flow works only with React pages.

The Claude Code artifacts docs are more specific than the launch tweet: artifacts can stay private, be shared inside an org, or be published to a public link. The Claude Tag security docs add that a Slack session can publish an artifact that remains available after the sandbox is released. The doc wrinkle: an older Help Center sharing page still says Team and Enterprise artifacts are org-only, while the updated Code docs describe org-owner enabled public links.

ClaudeDevs framed the upgrade as three changes, all aimed at making generated pages less trapped in a private chat, terminal session, or Slack thread.

• Public sharing: anyone with the link can view a published artifact, per ClaudeDevs' public-link note.
• Multiplayer editing: multiple people can edit the same artifact together on Team and Enterprise plans, per ClaudeDevs' launch thread.
• Claude Tag creation: a Slack thread can ask for a dashboard and get a working page back inside the organization, per ClaudeDevs' launch thread.

The earlier June Claude Code announcement described artifacts as live pages for PR walkthroughs, system explainers, dashboards, and release checklists that update as Claude Code works. This update changes the sharing and collaboration surface around that page.

The access model now has more knobs than the original Team and Enterprise beta. The Claude Code artifacts docs describe this setup:

• New artifacts are private by default.
• Pro and Max users can share through public links.
• Team and Enterprise users can share with specific people or the whole organization.
• Team and Enterprise public sharing starts off until an Owner enables it.
• The artifact header names the author and links to that author’s artifact gallery.

The Help Center sharing article still uses older app wording: publishing is listed for Free, Pro, and Max, while Team and Enterprise sharing is described as organization-only.

Multiplayer editing is URL-mediated. In the Claude Code artifacts docs, people shared onto an artifact are viewers by default; on Team and Enterprise, they can be switched to editor.

An editor updates the artifact by giving Claude the artifact URL in their own session. Claude pulls the current content, republishes a new version, and everyone with the page open sees the update live.

That matches trq212's workflow: a Claude Tag dashboard can be edited by collaborators or by local Claude Code sessions.

Claude Tag is the Slack side of the loop. The Claude Tag docs say a tagged task starts a working session in an Anthropic-hosted ephemeral sandbox, posts progress in the thread, and lets anyone in the channel steer the running task by replying.

The security docs add the artifact-specific rule: a session can publish a claude.ai page with the link posted in the thread, and anyone with access to the source Slack channel can open it. Someone without that channel access sees a request-access prompt instead of the page, according to Claude Tag security docs.

Public links do not turn artifacts into hosted apps. ClaudeDevs' public-link note frames the feature around prototypes, and the Claude Code artifacts docs draw hard boundaries:

• One self-contained HTML page.
• No backend.
• No stored form input.
• No API calls, fetch, XHR, or WebSocket calls at view time.
• No external scripts, stylesheets, fonts, or images.
• No multi-route app structure.

The docs position artifacts as a capture of work: annotated diffs, dashboards, option grids, timelines, and decision surfaces built from session context.

Before the launch, petergyang described a creator workflow that explains the demand for native collaboration: HTML artifacts as specs and plans, because they are easier to scan than Markdown files.

His requested interaction model was concrete:

• Select text or objects inside the artifact.
• Send that selection as feedback to the AI.
• Edit the artifact directly.
• Add comments and collaboration features inside the Claude Code or Codex web-viewer harness.

The attached artifact was a tabbed design-kit page with color tokens and source-of-truth notes. In the same thread, petergyang's reply said the existing edit button only worked with React pages, not HTML pages.