Can Claude Artifacts store data?

What artifact storage actually gives you

Anthropic documents this precisely, so here it is in their words rather than ours.

Persistent storage for artifacts is available on Pro, Max, Team, and Enterprise plans on Claude web and desktop.

Persistent storage for artifacts, Claude Help Center

• 20 MB per artifact. The help centre states a 20 MB storage limit per artifact, which is generous for records and small for anything else.
• Text only. The same page is explicit that there is no images, files, or binary data. A photo, a PDF or a spreadsheet upload is out of scope.
• Two pools. Personal storage: Each user maintains their own private data, while Shared storage: All users see and interact with the same data. Anthropic adds that Personal and shared storage are isolated.
• Published artifacts only. The condition below, which is the one that generates most of the confusion.

Why your storage code does nothing while you are testing

This is the single most common reason someone concludes that artifact storage is broken. It is not broken. It is off.

Persistent storage is only available for published artifacts. During development and testing, storage operations will not succeed until the artifact is published.

Persistent storage for artifacts, Claude Help Center

So the loop most people run, which is to build the artifact in the conversation, click around it in the preview pane, see nothing persist and assume the code is wrong, will fail every single time no matter how correct the code is. Nothing throws. Nothing warns. You have to publish first, open the published link, and test there.

Two practical consequences worth planning around:

• Your test data and your real data live in the same place, because there is only one place. There is no staging copy of a published artifact.
• Debugging happens on the live URL, which is also the URL you have given other people. Anyone holding the link is watching you work.

Personal storage versus shared storage

Choosing wrongly here is the difference between an app that works and one that leaks. The two pools do different jobs and neither one does the job most people actually want.

The gap between those two columns is where most real requests fall. A chore rota, a leave request form, a club membership list and a team tracker all need the same thing: one shared dataset where the app knows which person is looking. Artifacts do not offer that combination, and no amount of clever code inside the artifact can add it, because the page has no way to authenticate anyone. Anthropic says so directly:

An artifact is a static page. It can't store data submitted through a form or authenticate viewers itself.

Share session output as artifacts, Claude Code documentation

When artifact storage is the right answer, honestly

Plenty of the time it is, and this is worth saying because most pages on this topic will not say it. Keep using artifact storage when:

• The state belongs to one person. A calculator that remembers your last inputs, a reading tracker only you use, a form that survives a refresh. Personal storage is exactly this and costs you nothing to set up.
• The shared data is genuinely public. A vote tally, a guestbook, a collaborative word list. If it would be fine on a whiteboard in a corridor, shared storage is fine.
• It is going to be short-lived. Something you need for a weekend, a workshop or one sprint. Nothing to provision, nothing to bill, nothing to remember to delete.
• You want zero friction for the reader. There is no account wall on a published artifact at all, which is lighter than anything account-based can ever be, including us.

Against that, an artifact is the only tool in this comparison you can produce entirely inside a conversation, in seconds, with no deploy step. That is a real advantage and it does not stop being real because the storage has limits.

Signs you have outgrown artifact storage

The wall is usually one of five things. If you recognise two or more, you are not going to code your way past it:

• You need to know who did something, and shared storage has no author field to give you.
• Some rows should be visible to some people and not others. There is no permission model to express that.
• You need a file, an image or an attachment, and storage is text only.
• You are approaching 20 MB, or you would rather not think about a ceiling at all.
• The thing has quietly become important, and it currently lives at a URL that permanently deletes its own data if anyone unpublishes it. See what happens when you unpublish.

What changes if you move it to a hosted app

Homespun is where the same agent that built the thing can host it. People sign in with their email, so the app can tell who is who and every row remembers who created it; permissions are set per collection, so some people can approve and others can only submit; and the agent keeps read and write access to the collections it is allowed, so it can keep working on the data instead of handing you a page and walking away. It is free today, and there is no paid plan to buy.

Here is the honest comparison, including the columns where moving costs you something.

That last row is not a formality. If what you have is a thing for strangers to open once, an artifact is genuinely the better product and you should stay. The trade only pays off when the people using it are the same people every week.

How to move an artifact's data to a real app

The order matters, because of the silent-failure rule above and because unpublishing destroys the data.

• Get the data out while it is still published. Ask Claude, in the same conversation, to add a temporary export control to the artifact that reads everything out of storage and renders it as JSON in a textarea. Republish, open the published link, copy it. Storage only answers on the published page, so this cannot be done from the preview.
• Keep the code. The artifact's source is in the conversation. You will want it as the starting point rather than describing the whole thing again from scratch.
• Have the agent rebuild it as a hosted app. Connect Homespun to Claude, paste the artifact code and the exported JSON, and ask for the same app with real accounts and the data loaded. See connecting your AI.
• Only then unpublish, if you even want to. There is no rush, and unpublishing is permanent and cannot be reversed.

Questions people ask next

Primary sources

Every claim about Claude on this page comes from Anthropic's own documentation, quoted rather than summarised so you can check it yourself.

• Persistent storage for artifacts, Claude Help Center
• Publish and share artifacts, Claude Help Center
• Share session output as artifacts, Claude Code documentation