TL;DR
In July 2026 Claude shipped two things people had been asking for: public artifact links (anyone with the link can open a standard HTML artifact, no account needed) and real-time multiplayer editing (teammates edit one artifact together, with version history and rollback). Both are genuine upgrades. The catch is in how they're split: the public path is single-player, the multiplayer path is locked to your organization, and you can't have both at once. There's still no comment channel on the page, it only works on artifacts Claude itself built, and everything lives on claude.ai. If your work needs public and collaborative, with people who don't hold a Claude seat, that's the line this post is about.
What Claude actually shipped
Two features landed close together, and it's worth keeping them separate because their rules are different:
- Public sharing. Publish a standard HTML artifact and anyone with the link can open and interact with it, no Claude account required. This is the individual-creator path, and it's a real "send it to the world" link, the fastest one any AI tool ships.
- Multiplayer editing. On Team and Enterprise plans (in beta as this shipped), several people edit the same artifact in real time. When Claude produces a new iteration the open page updates in place and holds your scroll position, every revision is kept in version history, and a team can roll back. Artifacts can also be spun up from Claude Tag inside Slack.
For the jobs they're built for, both are good. Watching a build update live for your team beats pasting screenshots into a channel. A quick public link beats emailing a zip. None of what follows is "the feature is bad." It's about where the two roads stop.
The part that matters: the two halves don't meet
Here's the design decision underneath the launch. Read it as a map of where Claude will and won't go:
| You want | Public link | Multiplayer editing |
|---|---|---|
| Anyone with the link can open it | Yes | No, org members only |
| Several people edit at once | No, last save wins | Yes |
| The people editing need a Claude seat | To edit, yes | Yes, and in your org |
| It can be made public | Yes | No, org-scoped by design |
The public artifact is a broadcast: many can watch, one hand is on the keyboard. The multiplayer artifact is a room: many hands, but the door only opens for people already inside your company. There is no setting that is public and multiplayer and open to an outsider. That's not a gap waiting on a patch; it's the natural shape of a feature whose center of gravity is your own account and your own org.
The walls it kept
Past the split, four limits stay exactly where they were before July:
- No comments on the page. There's still no way for a viewer to pin feedback to the exact element they mean. Reactions come back as prose in chat, Slack, or a meeting, and someone translates them into prompts by hand.
- No editing for outsiders. A client, a prospect, or a contractor with no Claude seat can view a public artifact, but can't touch a word of it and can't join the multiplayer room.
- Claude-generated only. The whole thing works on artifacts Claude built. A ChatGPT canvas, a Gemini app, a v0 or Lovable export, or a page you wrote by hand doesn't get any of this. Real teams use more than one tool, and the output has to live somewhere neutral.
- Claude's domain, not yours. The link is a claude.ai URL. There's no custom domain and no way to put a client deliverable under your own brand.
Why this is good news, not bad
It's tempting to read a big-platform launch as "they're coming for the layer on top." The fine print says the opposite. Claude just drew the boundary of its collaboration story publicly, in its own product, and drew it exactly where a maker's tool naturally stops: inside the account, inside the org, on one tool's output, on one company's domain, with feedback still routing through chat.
The collaboration a mixed-tool team actually needs lives on the other side of every one of those walls: public and multiplayer at the same time, open to people who hold no seat in any AI tool, with comments pinned to the page, across whatever tool made the thing, on your own domain. A model-maker won't build that, because two of its requirements (be neutral across rival tools, and let non-customers in) run against the reason the feature exists at all.
How Coedit fits
Keep building in Claude, and use the new sharing for what it's great at: a fast public look, or a live build your teammates watch. When the work needs to be both open and worked-on, by people who don't live in your Claude account, that's the handoff. Export the artifact's HTML (it's a plain web page, and it's yours) and paste it into Coedit. You get one link anyone can open with no account, real multiplayer editing that doesn't care whose org they're in, comments pinned to the exact element, no-code editing for the people you allow, and one version history across every human and agent change, with rollback. On a paid plan it carries your own domain. And it works the same whether the page came from Claude, ChatGPT, Gemini, v0, Lovable, or your own hands. The merge engine behind the shared editing passes 47/47 concurrency tests.
Claude is the best place to make the thing. Coedit is where it meets everyone the maker's tool leaves outside.
FAQ
Q: Can I share a Claude artifact publicly now? A: Yes. Publishing a standard HTML artifact gives you a link anyone can open and interact with, no Claude account required. That public path is single-player, though: it's for showing the result, not for co-editing it with the viewer.
Q: Can two people edit the same Claude artifact at the same time? A: Yes, on Team and Enterprise plans, in real time with version history. The catch is that multiplayer artifacts are scoped to your organization, they aren't the public link. Someone outside your Anthropic org can't join the editing room.
Q: Can a client who doesn't use Claude edit or comment on my artifact? A: No. They can view a public artifact, but editing needs a Claude seat, multiplayer needs a seat in your org, and there's no comment channel on the page for them to use. Reaching people with no seat is exactly what a tool-agnostic layer like Coedit is for.
Q: Does the new sharing work on things I built in ChatGPT, Gemini, v0, or Lovable? A: No. It's for artifacts Claude generated. To collaborate on output from any tool (or a mix of them) on one surface, move the HTML onto a tool-agnostic layer.
Q: Can I put a shared Claude artifact on my own domain? A: No. Artifacts live at a claude.ai URL. For a client deliverable under your own brand, publish the page through a tool that supports custom domains.