← All posts Compare

Claude Code Artifacts: sharing limits and what to do instead

TL;DR

Claude Code Artifacts (beta, launched June 18, 2026, on Team and Enterprise plans) turns a Claude Code session into a self-contained, live HTML page at a URL that updates as the session keeps working. A July 2026 update added public links and real-time multiplayer editing, so the "org members only" wall is no longer absolute. But the two are split: the public link is for showing the result, and the multiplayer editing stays scoped to your organization. Even after the update there's still no comment channel on the page, no editing for anyone outside your org, no custom domain, and it only works on what Claude itself built. If the person you need to reach is a client or a contractor with no Claude seat and you need them to do more than look, that's the gap this post covers.

What Claude Code Artifacts actually is

Anthropic describes it as a single, self-contained interactive HTML page published from a Claude Code session to a private URL on claude.ai, rendering in any browser and updating in place as the session continues: PR walkthroughs, system diagrams, dashboards, release checklists, the kind of thing that used to live only inside a terminal or a screen share. Boris Cherny, who leads Claude Code at Anthropic, called it "a game changer for how I work with Claude." For the job it's built for, watching a session's work update live for people already on your team, it's a real upgrade over pasting screenshots into Slack.

It's a different product from the artifact feature in the regular Claude.ai chat (what is a Claude artifact, explained here): that one is a chat-published output anyone can generally view with a link; Claude Code Artifacts is scoped to a coding session and an organization.

The walls that are left

The July 2026 update softened the biggest one: you can now publish a public link, and teammates can co-edit in real time. But the update kept the two apart (public or multiplayer, never both at once), and the walls that matter for outside work all stayed up:

  • The public link is single-player. Anyone can open a published artifact, but it's for viewing the result. The real-time co-editing is a separate, org-scoped mode.
  • Multiplayer stays inside your org. Collaborative editing is Team/Enterprise and scoped to your organization. A client or contractor outside your Anthropic org can't join the room.
  • No comments pinned to the page. Feedback still routes back through chat, Slack, or a meeting; there's no way for a viewer to mark up the exact element they mean.
  • No editing for outsiders. Someone with no Claude seat can view a public artifact but can't change a word, and can't be invited into the multiplayer room.
  • No custom domain. The link lives on claude.ai, so a client deliverable can't carry your own brand.

That's the same shape as Codex Sites, OpenAI's equivalent feature, and it's the same shape the public-vs-multiplayer split draws across the whole artifacts launch: the model-makers build "share this build with my team," and stop at the company's edge. It's a sensible scope for an internal-tooling feature. It's still a wall the moment your actual audience is outside it, or needs to do more than look.

Comparison at a glance

Claude Code Artifacts Codex Sites Coedit
Builds the app for you Yes (from a Claude Code session) Yes (from a prompt) No, Coedit doesn't generate
Public "anyone with the link" Yes, but view-only (single-player) No Yes
Real-time multiplayer editing Yes, but org-only (not public) No Yes, and public
Editing for people outside your org No No Yes
Custom domain No No Yes, on paid plans
Comments on the live page No No Yes
Non-coder editing No No Yes
Works across tools (Claude, ChatGPT, Gemini, v0, Lovable, hand-written HTML) Claude Code sessions only Codex output only Yes

How Coedit fits

Keep building in Claude Code (or anything else). When the artifact needs to reach someone outside your Anthropic org, a client, a prospect, a contractor without a seat, export the page's HTML and paste it into Coedit. You get one link anyone can open with no account, comments pinned to the exact element, no-code editing for people you allow, and version history with rollback, on your own domain if you're on a paid plan. Claude Code Artifacts is the best way to watch a session build for your team. Coedit is the way to hand the result to everyone else.

FAQ

Q: Can I share a Claude Code Artifact with someone outside my company? A: They can view a public link, since the July 2026 update. But they can't edit it or comment on it: the real-time multiplayer editing is scoped to your Anthropic org, and there's no comment channel on the page. So an outsider can look, not work with you. To let a client or contractor without a Claude seat actually collaborate, move the page onto a tool built for outside collaboration.

Q: Does Claude Code Artifacts support a custom domain? A: No. Artifacts are hosted at a claude.ai URL with no option to serve them from your own domain.

Q: How is this different from a regular Claude artifact? A: The chat-based Claude.ai artifact is published from a conversation and generally viewable via a link (details here). Claude Code Artifacts is a separate feature scoped to a Claude Code session; the July 2026 update added public links and org-scoped multiplayer editing to it (full breakdown here).

Q: What's the workaround to share a Claude Code Artifact externally? A: Export or copy the page's HTML and publish it through a tool built for external sharing, such as Coedit, which adds a public link, comments, no-code editing, and version history on top of whatever built the page.

Your AI work shouldn't stop at a file.

Turn the page your AI made into a link anyone can open, comment on, and edit. No code, no account to view.

Get your live link →