Anthropic pushed out a meaningful batch of Claude Code updates this week that are worth knowing about, particularly if you’re using it for anything beyond basic code generation. Here’s what’s new and what it actually means in practice.
In-App Browsing — Claude Can Now See External Websites
The Claude Code desktop app now has a built-in browser. Open it with Ctrl+Shift+B on Windows (Cmd+Shift+B on macOS), or find it under the Views menu.
What makes this more than a curiosity: Claude can read, click, and interact with websites inside this browser the same way it already handles your local dev servers. So if you’re building something that needs to talk to an external site — a scraper, a form integration, a checkout flow — Claude can now open the live page, inspect what’s actually there, and work with it directly rather than reasoning from a description you’ve typed out.
It’s sandboxed and configurable, so you decide whether sessions persist. If you’ve ever spent time explaining to an AI what a third-party UI looks like when you could have just shown it, this closes that gap.
Artifacts — Now With Live MCP Data
Artifacts have been in Claude Code for a while — they’re the interactive pages Claude builds from your session, like a PR walkthrough or a project dashboard. They publish to claude.ai, stay private to your account by default, and can be shared via a public link when you want them to be.
The notable new capability: artifacts can now call MCP connectors directly. That means an artifact isn’t just a static snapshot of your session anymore — it can pull live information and take actions for whoever’s viewing it, on demand.
The practical implication is significant. You could build an artifact that lets a client or teammate check the current status of something, trigger a workflow, or query a data source — without them needing access to Claude Code themselves. The artifact becomes a lightweight app, not just a document.
iOS Simulator Support (Beta, macOS Only)
If you’re doing any iOS development, Claude Code can now build your app, launch the iOS simulator, and verify the result — all without leaving the session. It requires Xcode and is currently macOS only.
The value here is the same as the in-app browser: Claude can see what it built and confirm it actually works, rather than handing you code and leaving verification to you. For anyone doing iOS work, it’s a meaningful step toward a tighter loop between generation and validation.
Linux Desktop App (Beta)
The Claude desktop app is now available on Ubuntu and Debian. Until now, Linux users were limited to the CLI — now they get the full desktop experience, including running sessions in parallel with Git isolation, monitoring CI with auto-fix and auto-merge, and reviewing diffs inline.
One thing worth highlighting for anyone managing multiple environments: the desktop app shares the same configuration as the CLI. Your CLAUDE.md files, MCP servers, and skills carry over automatically. No separate setup required.
The Bigger Picture
What ties these updates together is Claude Code moving from a tool that operates on files to a tool that can engage with the full context of what you’re building — live websites, live data, running simulators, multiple platforms. Each addition shrinks the gap between “Claude wrote the code” and “Claude verified it works in the real environment.”
If you’ve been using Claude Code mainly for code generation within a single codebase, it’s worth revisiting what it can handle now. The capability set has shifted considerably in the past few months.
Using Claude Code in your business or IT workflow? Get in touch — always happy to talk through what’s practical for your situation.