Anthropic announced a new feature that could be a game-changer for Claude Code users on the Claude Team and Enterprise subscription plans: Artifacts.
This update turns the work of a Claude Code session into a custom, interactive, shareable HTML web page, allowing a Claude Code user to connect live code, multiple data sources, and display it in an interactive URL that they can send to other teammates, whether it’s a dashboard, an app design, or some other product intended for internal use.
These teammates and the original user can see the web page updating in real time as Claude Code performs its work autonomously or under the user’s guidance, and as data sources and connected code bases change.
While Anthropic first introduced Artifacts in its consumer web chatbot in the summer of 2024, where it evolved from a manual toggle feature to a generally available tool for publishing code snippets and games to the web, integrating this capability directly into Claude Code’s command-line interface (CLI) and desktop app bridges the gap between deep back-end engineering and the non-technical stakeholders who need to understand it.
Product and technology: the end of the status update
In essence, Claude Code Artifacts acts as a dynamic translation layer. Built directly from the uninterrupted context of a user’s session, the agent uses the local repository’s code base, connected monitoring tools, and conversational reasoning to create specialized web pages.
Engineers no longer need to connect external data sources or install temporary infrastructure; AI builds the user interface from what already exists.
Crucially, these web pages are not static exports. As the AI works through a terminal session, the open web page is updated in place, updating graphics and text instantly at the exact same URL. Each update publishes a new version history, allowing teammates to roll back or track agent progress securely on a desktop or mobile device.
The Battle of Live, Interactive, Shared AI Worksurfaces: Anthropic’s Claude Code Artifacts vs. OpenAI’s Codex Sites
Anthropic update arrives more than two weeks after OpenAI released a massive update to its own Codex platformintroducing a strikingly similar business hosting feature called "Sites".
This tit-for-tat product cadence highlights a rapidly escalating battle in the enterprise workspace between functions and beyond the developers themselves, although there are some important technical and philosophical distinctions worth noting for companies considering either.
As revealed on their respective developer documentation web pages, OpenAI is building a platform as a service; anthropic you are building a stateless canvas.
OpenAI Sites is designed to generate durable and complete web applications. According to the platform documentation, Codex Sites hosts projects that are built as Cloudflare Worker-compatible ES modules.
Fundamentally, Sites supports a persistent backend infrastructure: agents can connect automatically "D1" relational databases for structured data (such as user progress or saved records) and "R2" Object storage for file uploads. An OpenAI site can support public logins, integrate with third-party identity providers, and allow very specific access controls tailored to specific workspace groups.
It uses a two-stage release process: saving a reviewable candidate linked to a Git commit before officially deploying it to production. In short, it is a production environment designed to replace functional internal SaaS tools.
Anthropic’s Claude Code Artifacts, on the other hand, deliberately avoids the backend. The newly published documentation is blunt about its limitations: "An artifact is a capture of work, not an application.".
Each artifact is a single, independent HTML page with a rendering size limit of 16 MiB. To ensure the security of the organization, Claude wraps the published file in a strict Content Security Policy (CSP) that blocks all requests from the external network. t
This means that the page cannot load external scripts, fonts, or style sheets, and fetchXHR and WebSocket calls are completely blocked. All CSS and JavaScript must be integrated and images must be embedded as data URIs. Artifacts cannot store form inputs, call an API at view time, or serve multiple paths.
This technical limitation is actually Anthropic’s deliberate philosophical position: while OpenAI wants to create enterprise-wide persistent software portals, Anthropic keeps Claude Code firmly anchored in highly secure, ephemeral technical workflows. Claude’s artifacts are No intended to be software; They are intended to replace whiteboard diagrams, manual bug tours, and status reports with secure, self-updating visual tools that never leak live data outside corporate boundaries.
Enterprise Security and Licensing: Keep the code base private
Because these agents sit at the nexus between company proprietary data and live codebases, licensing and access controls are a primary concern.
Both Anthropic and OpenAI have opted for closed, proprietary licensing models for these new visual workspaces. For end users and developers, the distinction is critical. Unlike permissive open source software (such as MIT or Apache 2.0) or strict copyleft licenses (such as GPL), which give developers the legal freedom to inspect, modify, and host the underlying code, neither Claude Code Artifacts nor Codex Sites can be forked or hosted independently.
Enterprise customers do not retain code-level ownership of the Anthropic rendering engine or Codex integration nodes; Both operate strictly within their infrastructures managed by the respective creators.
To make this vendor-managed approach acceptable to enterprise compliance teams, both companies have heavily prioritized organizational security. Anthropic ensures that each artifact is private to its author by default and strictly cannot be made public on the Internet at large. When an engineer chooses to share a link, it can be viewed only by authenticated members of their specific organization. System administrators retain ultimate authority, managing access through organization-level toggles, role-based scope, and explicit retention policies, while maintaining oversight through a centralized compliance API.
OpenAI takes a similar closed approach with Codex Sites, rolling out the feature primarily for the ChatGPT Business and Enterprise workspaces. Like Anthropic, OpenAI relies on system administrators to manage deployment through centralized workspace configurations, requiring an administrator to explicitly enable sites through role-based access control (RBAC) for enterprise levels.
However, because Codex Sites works more like a hosted web application, its access controls are slightly more granular. When an engineer prepares to share a deployed URL, they can apply specific access modes: restrict the site to only themselves and workspace administrators, open it to all active users in the workspace, or limit access to custom user groups.
Additionally, to prevent sensitive data leaks, OpenAI provides a dedicated Sites panel to manage runtime variables and secrets securely, ensuring those keys don’t have to be committed to local source files.
Reactions and Reflections
The introduction of visual, self-updating UI layers in command-line agents is fundamentally altering the way developers view their own workflows. As AI handles raw syntax and automates reporting, the friction of communicating technical work to stakeholders is disappearing.
Boris Cherny, leader and creator of Claude Code, highlighted the sheer usefulness of the update in a post on X earlier today:
"I’ve been using Artifacts in Claude Code for everything: visual explanations of complicated code, system diagrams, quick previews of some animation options, data analysis, and dashboards that I share with the team." Cherny wrote. "They are a turning point in the way I work with Claude. I can’t wait to hear what you think!"
This sentiment is practically demonstrated in Anthropic’s launch materials. In one scenario, an engineer asks Claude Code to investigate user churns from a previous version of the software.
In a matter of seconds, the agent runs a SQL read, creates an interactive delivery funnel dashboard, and diagnoses that "Pro accounts stuck on export sheet". The AI then proposes UI fixes, updates live graphics as code is refactored, and generates a secure link that an administrator can instantly open via mobile device.
By turning the terminal into a live, collaborative canvas, Anthropic is demonstrating that the most valuable output of an AI coding assistant is not just the code itself: it’s the context, the reasoning, and the ability to share that work instantly.
