Anthropic Brings Live, Shareable Artifacts to Claude Code

Anthropic Brings Live, Shareable Artifacts to Claude Code

Anthropic adds live, shareable Artifacts to Claude Code, allowing teams to view AI coding sessions as real-time, interactive pages.

Anthropic this week extended its Artifacts feature into Claude Code, giving engineering teams a way to turn an AI coding session into a live web page that colleagues can open, explore, and watch updates in real time. The feature is in beta for Claude Team and Enterprise subscribers and works from both the Claude Code CLI and desktop app.

It addresses a familiar gap in how engineering work gets shared. A developer running Claude Code through an incident investigation, a service refactor, or a multi-month data analysis usually has two options for keeping teammates in the loop: wait until the work is done, or send a screenshot and a written update. Artifacts is meant to close that gap by turning the session itself into something teammates can open directly.

How it Works

An artifact is built from the full context of a Claude Code session, including the local codebase, any connected tools or plugins, and the conversation that produced the work. The result is a single, self-contained interactive HTML page, complete with charts, forms, and other interface elements, hosted at a private link.

Anthropic’s documentation is specific about what an artifact is not. It’s a capture of work, not an application. Each page is capped at 16 MiB and runs under a strict content security policy that blocks external network requests, meaning no outside scripts, fonts, or stylesheets, and no live API calls. All CSS and JavaScript is inlined, and images are embedded directly. Artifacts can’t accept form submissions or serve multiple routes. That’s a deliberate constraint: it keeps the feature scoped to sharing a snapshot of work rather than standing up a backend service inside an enterprise network.

What makes it more useful than a static export is the live update behavior. When Claude Code produces a new iteration, the open page refreshes in place and holds the viewer’s scroll position. Teammates see updates the moment they’re published, and every revision is kept in version history, so a team can roll back to an earlier version if needed.

Access defaults to private. Every artifact is visible only to its author unless shared, and sharing is limited to authenticated members of the same organization. Admins can manage access through roles and retention policies, which matters for teams that need to keep generated pages, especially anything touching production data, inside normal governance boundaries.

Anthropic points to a fairly wide set of use cases:

• Pull request walkthroughs
• Incident timelines
• License audits
• Architecture overviews
• Anything else where a person would otherwise have written a status update by hand

Where This Fits in Claude Code’s Trajectory

The launch comes as Claude Code’s enterprise footprint continues to expand. The product became generally available in May 2025 and has scaled quickly since, with enterprise customers now accounting for more than half of its revenue. Anthropic’s broader run-rate revenue crossed $47 billion in late May 2026, up from roughly $9 billion at the end of 2025, a jump driven largely by enterprise adoption and Claude Code specifically. That figure came alongside the company’s $65 billion Series H round, not as part of this artifacts announcement, but it gives some sense of the scale Claude Code is now operating at inside large organizations.

Mitch Ashley, VP and practice lead for software lifecycle engineering and AI-native software engineering at The Futurum Group, sees the move as part of a larger shift in what teams actually scrutinize about AI coding tools.

“The contested layer in AI coding tools is moving from code generation to the surface where teams inspect and trust an agent’s work,” Ashley said. “A governed, org-scoped record of the session, with access controls and retention, makes that visibility deployable inside an enterprise.”

He added that the format itself doesn’t settle the question of trust on its own.

“Engineering and security teams weigh these tools on the visibility and control they hold over agent work,” he said. “A live page that reads as authoritative still has to earn it through review, or it scales the trust-without-verification it appears to resolve.”

The Bigger Pattern

Artifacts in Claude Code is a small feature with a clear logic behind it: AI-assisted engineering work increasingly happens in long sessions that span hours or days, and the people who need visibility into that work, whether that’s a manager, a security reviewer, or another engineer picking up where someone left off, shouldn’t have to wait for a final write-up to see it.

It also reflects something broader happening across AI coding tools generally. As these tools take on more of the implementation work, the bottleneck shifts from writing code to explaining what the code does and why. Anthropic’s bet here is that a live, sandboxed web page is a better communication format than a chat transcript or a status message, particularly for work that’s still in motion.

The security tradeoffs built into the feature - no external calls, no persistent backend, strict size limits - suggest Anthropic is trying to make this safe enough for security-conscious enterprises to turn on without much hesitation. Whether that’s enough to make Artifacts a regular part of engineering workflows, rather than a nice demo, will depend on how teams actually use it once the beta opens up more broadly.