For the past couple of years, our relationship with artificial intelligence has felt a bit like managing an enthusiastic but literal-minded intern. You type a prompt, you get a response, you type another prompt, and then you spend ten minutes fixing the formatting. It’s a constant, active back-and-forth conversation trapped inside a familiar chat box.

When Anthropic introduced Claude Cowork in January 2026, it was a deliberate attempt to break out of that conversational bubble. The idea was simple: instead of chatting with an AI, you handed it an entire project, walked away, and let it execute multi-step workflows natively. But as powerful as it was to watch an agent sort through chaotic local directories or synthesize dense stacks of documents, it suffered from a frustrating, old-school limitation. It was tethered entirely to your physical desktop. If you closed your laptop to run to a meeting, grab a coffee, or catch a flight, the work ground to a halt.

That friction point is officially dissolving. In a significant product shift, Anthropic announced that Claude Cowork is coming to web and mobile, transitioning the feature from a locked-down desktop app to a cross-device cloud experience. It is a subtle but massive architectural pivot that hints at exactly where the broader AI agent race is heading.

By moving Cowork sessions onto Anthropic’s own servers for remote execution, the agent can now run fully in the background. You no longer need to keep your machine awake or stay glued to a desk. You can queue up a heavy data reconciliation task or request a comprehensive competitor brief, shut your laptop for the night, and let the cloud handle the heavy lifting. The update even introduces scheduled tasks that can execute when your devices are completely offline. Imagine setting a prompt to crawl internal email threads and draft a client briefing document at 6:00 am, so it’s sitting ready for your review the moment you log on with your morning coffee.

The real magic of the mobile expansion, however, lies in how it handles human oversight. One of the persistent anxieties of letting an autonomous agent loose on your workflows is the fear that it will go off the rails or stall out indefinitely when it encounters an ambiguous problem. Anthropic’s fix for this is clever: when Claude hits a crossroad or needs a crucial decision that only you can make, it pings a notification directly to your phone. You can steer a draft, approve a spreadsheet format, or answer a clarification question mid-commute, and the agent will immediately resume its path.

This hybrid model acknowledges a workplace reality that many tech companies overlook: desktop remains the place for deep, concentrated work, but the management of that work happens everywhere. Armmand Hosseini, a customer success professional at Ramp, noted during early testing that he was able to build a client-tracking dashboard on his laptop and seamlessly pick the session back up on his phone while waiting at an airport baggage carousel. The thread never dropped.

What makes this expansion particularly interesting is who is actually using these tools. When AI agents first started gaining traction, they were almost exclusively the playground of software engineers writing and debugging code. But Anthropic’s internal data revealed a fascinating trend: more than 90% of Cowork usage has nothing to do with software development. Instead, it’s being dominated by everyday knowledge workers in business operations and content creation.

It turns out people are using the tool to tackle what tech teams call “the work around the work”—those time-consuming, repetitive tasks like translating a messy folder of client contracts into a neatly structured renewals tracker with risks flagged, or building slide decks out of raw call transcripts. It’s the administrative scaffolding that clogs up everyone’s calendar but rarely shows up on a formal job description.

By launching a web version alongside the mobile apps, Anthropic is also democratizing access for corporate environments where strict IT policies prohibit downloading local desktop software. For users who were previously locked out by security firewalls, the web dashboard offers a single, unified home where standard chat, projects, and autonomous Cowork sessions live together.

The rollout is beginning as a gradual beta over the next several weeks, starting with Claude Max subscribers before expanding to Pro and other tiers. To sweeten the deal and encourage people to throw larger, more complex tasks at the system, Anthropic is doubling Cowork usage limits through August 5.

Ultimately, this move represents a broader philosophical shift in personal computing. We are moving away from an era where we use software to do our jobs, and toward an era where we manage AI systems that execute the jobs for us. By untethering the agent from the desktop and letting it run quietly in the cloud, Anthropic is making a strong case that your most productive colleague doesn’t need an office desk—just a reliable server connection and a way to text you when it gets stuck.