What's the difference between a Claude live artifact and a regular Claude artifact?

A regular artifact is a snapshot — Claude generates it once and it's frozen at that point. A live artifact (in Claude Cowork) refreshes from your connected data sources through MCP every time you open it, so the numbers reflect the current state. Live artifacts also live in a dedicated sidebar tab rather than inside a chat thread.

Can I share a Claude live artifact with someone who doesn't have Claude?

Not yet. As of April 2026, live artifacts are device-local and can't be shared. Sharing is on Anthropic's roadmap. For now, the workaround is to export the HTML and send the file, though any MCP-powered data refresh won't work for the recipient. Regular (non-live) artifacts can be published as a public link anyone can open.

Do I need to know how to code to use Claude live artifacts?

No. You describe what you want in natural language — 'build me a tracker that pulls open tickets from Linear and shows them by priority' — and Claude writes the code and MCP calls. You don't see or edit the HTML unless you want to.

What plan do I need for Claude live artifacts?

Claude Pro ($20/month) or higher — Max, Team, or Enterprise. Free accounts don't have access to live artifacts or the Cowork interface. You also need Claude Desktop (macOS or Windows), not the browser version.

Do Claude live artifacts update automatically, or only when I open them?

Only when you open them. The artifact makes its MCP calls on load — it doesn't poll in the background or update while your laptop is closed. If you need something that monitors continuously (alerts, scheduled reports), a live artifact isn't the right tool for that part of the workflow.

Can stakeholders leave comments on a Claude live artifact?

No — Claude has no native comment layer, live or otherwise. To collect pointed feedback (someone clicking a specific element and leaving a note), you need to publish the artifact to a tool that supports that. Drafty is one option: publish there and anyone can click the exact element and comment without an account.

drafty

Claude live artifacts — what they are and what they can't do yet

A live artifact is a persistent, data-connected dashboard Claude builds for you in Cowork. It pulls from your apps when you open it so the numbers reflect today, not last Tuesday. The catch: it lives on your laptop and you can't share it yet.

Quick answer

A Claude live artifact is a persistent HTML dashboard that Claude creates in Claude Cowork (Desktop app). Unlike a regular artifact — which is a snapshot frozen at generation time — a live artifact refreshes from your connected data sources (Linear, Slack, Google Calendar via MCP) when you open it. Requires a paid Claude plan. Not yet shareable with others; sharing is on the roadmap.

Live artifacts vs regular artifacts

Most people use the word "artifact" for anything Claude renders in the preview panel — a landing page, a chart, a doc. That's the original artifact: a snapshot. You ask, Claude generates, the pane shows the result. Change the data source? Regenerate from scratch.

A live artifact is a different thing. It's a persistent, interactive HTML page that lives in your Claude Desktop sidebar under a dedicated "Live artifacts" tab. When you open it, it calls your connected apps through MCP and pulls fresh data — so a metrics dashboard shows this week's numbers, not the numbers from when you built it.

	Regular artifact	Live artifact
Where it lives	Inside the chat thread	Dedicated sidebar tab
Data	Snapshot at generation time	Refreshes from connected sources on open
Cross-device	Accessible in Claude.ai	Device-local (doesn't sync)
Shareable	Yes, via Publish button	Not yet (roadmap)
Plan required	Free + paid	Paid (Pro, Max, Team, Enterprise)
Requires Claude Desktop	No	Yes (macOS or Windows)

The pattern to remember: regular artifacts are for things you make and hand off. Live artifacts are for things you keep and return to.

What Claude Cowork live artifacts are actually good at

Live artifacts sit in a useful middle zone: above a spreadsheet in interactivity, below a full SaaS dashboard in complexity.

Weekly metrics dashboards. You describe what you want to track — conversion rate from your analytics tool, open tickets from Linear, Slack channel activity — and Claude wires up the MCP calls. Open the artifact on Monday morning and it shows Monday's numbers.

Project trackers. A live artifact can pull from Linear, your calendar, and Slack simultaneously. The result is a single-pane view of a project without toggling between three tools.

Competitive intelligence monitors. Combine an MCP connector to a web search or pricing API and the artifact shows the current state rather than the state at last week's meeting.

Morning briefings. Several teams use live artifacts as a personal summary page — open it at 9am and it consolidates your Slack mentions, today's calendar, and pending PRs.

The underlying mechanic: every time the artifact opens, it re-runs its MCP calls. You get a view that reflects now, not when you built it.

The gap that matters for makers and PMs

Live artifacts are built for the person who made them. They live on your computer. They don't sync to your phone or another laptop. And as of the Cowork launch (April 2026), they can't be shared with others — the recipient would need to recreate the MCP connections on their own machine to see anything useful.

This creates a real split in the workflow.

For internal, solo use — a dashboard you open yourself every morning — live artifacts are excellent. For any loop that involves another person — a stakeholder who needs to see the tracker, a client who wants to review the prototype, a PM who's not on Claude Desktop — the artifact stays stuck on your machine.

The workarounds people use:

Screenshot the artifact and paste it into Slack (loses interactivity, loses data freshness)
Export the HTML and send the file (requires the recipient to open it in a browser manually, and any MCP-dependent data won't load)
Share a screen recording (fine for one-off demos, terrible for async review)

None of these close the feedback loop. Stakeholders can't point at the specific row in the tracker and comment. You can't see what they reacted to. The iteration cycle that live artifacts enable for your own data-refresh loop breaks the moment you need someone else in it.

What you can do with a regular published artifact

Regular Claude artifacts — the kind you create in any Claude.ai conversation, not just Cowork — do have a Publish button. That gets you a link anyone can open without a Claude account. But published artifacts are static: there's no MCP connection, so what the viewer sees is the state at publish time. And there's no native way to leave a comment on the specific element.

For review workflows — showing a client a prototype, getting a stakeholder to sign off on a dashboard layout, collecting feedback on a PRD — the gap between "published link" and "pointed feedback" is where things slow down.

Where Drafty fits

Drafty is one option for that gap. You paste your artifact (or have Claude push it), and get a drafty.im/canvas/… link that renders on any device. Anyone can click the exact element and leave a comment — no account needed. Claude reads those comments through the CLI and ships an updated version on the same URL, with version history. It works on artifacts from any tool, not just Claude.

What you actually need to use Claude live artifacts

A few requirements that aren't obvious from the announcement:

Claude Desktop (macOS or Windows). Live artifacts don't exist in the browser version of Claude. You need the desktop app, updated to the latest version.

A paid plan. Pro ($20/month), Max, Team, or Enterprise. Free accounts don't get access to live artifacts or the Cowork interface.

MCP connectors set up. Live artifacts that pull from external sources (Linear, Slack, Google Calendar, databases) need those MCP connectors configured first. Without connectors, you can still build a live artifact, but it won't refresh external data — it's just a persistent interactive page.

Your data must be MCP-accessible. If your data lives in a manually-updated spreadsheet with no API, a regular artifact that you regenerate when needed is simpler.

A thing most guides get wrong

Most coverage of Claude live artifacts focuses on how to create them. The more useful frame for a maker or PM is: what's the right artifact for the right job?

Personal tracking, data you check solo: live artifact. The refresh-on-open mechanic is exactly right.
A thing you're handing to someone else: regular artifact, published to a link. Or a live artifact exported and hosted somewhere the recipient can reach without Claude Desktop.
Iterative review with stakeholders: neither artifact type has native comments. That step still needs something else — a review tool, a screenshot in Slack, or a shared canvas like Drafty.

The artifact pane (live or not) is the best make-and-iterate loop in any AI tool right now. The review loop is still a separate problem.

Claude live artifacts FAQ

What's the difference between a Claude live artifact and a regular Claude artifact?: A regular artifact is a snapshot — Claude generates it once and it's frozen at that point. A live artifact (in Claude Cowork) refreshes from your connected data sources through MCP every time you open it, so the numbers reflect the current state. Live artifacts also live in a dedicated sidebar tab rather than inside a chat thread.
Can I share a Claude live artifact with someone who doesn't have Claude?: Not yet. As of April 2026, live artifacts are device-local and can't be shared. Sharing is on Anthropic's roadmap. For now, the workaround is to export the HTML and send the file, though any MCP-powered data refresh won't work for the recipient. Regular (non-live) artifacts can be published as a public link anyone can open.
Do I need to know how to code to use Claude live artifacts?: No. You describe what you want in natural language — 'build me a tracker that pulls open tickets from Linear and shows them by priority' — and Claude writes the code and MCP calls. You don't see or edit the HTML unless you want to.
What plan do I need for Claude live artifacts?: Claude Pro ($20/month) or higher — Max, Team, or Enterprise. Free accounts don't have access to live artifacts or the Cowork interface. You also need Claude Desktop (macOS or Windows), not the browser version.
Do Claude live artifacts update automatically, or only when I open them?: Only when you open them. The artifact makes its MCP calls on load — it doesn't poll in the background or update while your laptop is closed. If you need something that monitors continuously (alerts, scheduled reports), a live artifact isn't the right tool for that part of the workflow.
Can stakeholders leave comments on a Claude live artifact?: No — Claude has no native comment layer, live or otherwise. To collect pointed feedback (someone clicking a specific element and leaving a note), you need to publish the artifact to a tool that supports that. Drafty is one option: publish there and anyone can click the exact element and comment without an account.