An artifact is one thing you and your agents work on together. It is not a file in a folder. It is a live app, a readable document, or a versioned file, with one address and one history.
That definition does the work. Because it is live, everyone who opens it sees the current version. Because it keeps a real history, an agent can edit it properly rather than regenerating it from scratch.
Why not just files
A file gets copied. Someone downloads it, edits it, sends it back, and now there are two. With agents in the loop that gets worse fast, because an agent will happily produce version four while you are reading version two.
An artifact cannot be copied by accident. There is one of it, at one URL, and every change lands in the same history. If a change was wrong, you restore an earlier version rather than hunting for an older copy.
What every artifact has
| Property | Meaning |
|---|---|
| Address | Its own URL, which never changes |
| Type | App, doc, or asset, which decides how it renders |
| Owner | The person who created or claimed it |
| Team | The team it belongs to |
| Visibility | Private to the team, or published |
| History | Every change, with an author and a message |
Its address looks like app.agentgrid.io/artifacts/Ez-CDDsmYLaApCIqpPv5bQ — the workspace, then the artifact's own id. Copy the real one from Share.
The three states
Anonymous. An AI that was never connected created it. Live and viewable by anyone with the link, but read-only, owned by nobody, and expiring 24 hours after creation.
Claimed. Someone took ownership. It belongs to a team, editing is unlocked, and the expiry is gone. Work from a connected assistant starts here directly.
Published. Deployed to a public URL that anyone can open with no account. Separate from sharing, and always an explicit step.
What can be an artifact
- An app. A running web page — a presentation, a dashboard, a calculator, a form, a game. Renders live at its URL, and its card in the hub shows a live preview thumbnail.
- A document. One or more markdown files, rendered as a readable page you can edit in the rich editor.
- Anything else. Stored as an asset: versioned, shareable, and readable by your agents, but with no preview. AgentGrid.io tells you plainly when it cannot render something rather than showing a broken frame.
What people make as apps
"App" sounds technical, and it is not. If it can be built with HTML, it can be an app here — anything that is not a document and benefits from being seen rather than read.
| At work | Studying and teaching | Personal |
|---|---|---|
| A slide deck or presentation | A quiz or flashcard set | A trip itinerary with a map |
| A dashboard or report with live charts | An interactive lesson or explainer | A budget or savings tracker |
| A pricing or ROI calculator | A revision timeline | A recipe collection |
| A landing page or one-pager | A lab-results write-up with charts | A wedding or event invitation |
| A project timeline or roadmap | A concept diagram you can click through | A personal portfolio or CV |
| A survey or intake form | A study planner | A countdown or checklist |
| A directory or comparison table | A citation and reading list | A photo gallery |
Presentations are worth calling out: a deck built as an app opens in a browser at one address, updates when it changes, and never needs to be re-sent as an attachment.
Limits
- An app needs an
index.html. Without one it renders as a blank page. - A document needs at least one
.mdfile. - App source changes through your AI or Edit with AI. Document content is also editable directly in the app.
FAQ
Is an artifact the same as a project? Roughly, but the word matters here. A project implies a folder you sync. An artifact is a single live thing with one address.
Can I put a whole codebase in one?
Yes. Ask your AI to push the project to AgentGrid.io; send source and assets, not node_modules or build output. The in-app Import from ZIP route is coming soon.
What happens if I delete an artifact? Everything goes with it, and it is not recoverable. That is different from restoring: restore rolls content back and keeps the history, delete removes all of it.
Can two agents work on the same artifact? Yes. The history records each of them separately, by name.