What happened
Anthropic just expanded Claude Cowork beyond the desktop, rolling it out to mobile and web so AI agents can keep working even when you close your laptop. Announced on July 7, 2026, the update means your Cowork sessions and files now follow you across devices instead of being tied to a single machine. Beta access is rolling out over the next several weeks, starting with Max plan subscribers, with broader plan availability to follow.
Cowork itself isn't new — it's the feature where you hand Claude an open-ended task and it works across your files, calendar, email, messaging apps, the web, and any other connected tools until the job is actually done, not just answered. What's new is the reach: start a task on your desktop, check progress from your phone on the train, and pick up the finished output wherever you happen to be.
Anthropic also disclosed usage data that reframes what "AI agent work" actually looks like in practice. Despite AI agents building their reputation on writing code, more than 90% of Cowork usage isn't software development at all. The two largest categories — business operations and content creation — together account for roughly half of all usage. Think reconciling quarterly spend into a variance memo, turning a folder of contracts into a renewals tracker with risks flagged, or building a client deck from call transcripts and pipeline data overnight.
### The core technical shift
Three specific changes make this possible: work now follows you across devices, tasks continue running in the background after you close the laptop, and scheduled tasks execute even with no device online. That last point matters most — a task set for 6 a.m. Monday genuinely runs on Anthropic's infrastructure, not on your idle laptop.
Why it matters
Until now, Cowork lived entirely on your laptop, which meant the moment you stepped away, the work stopped too. That's a real constraint for anything that takes longer than a single sitting — and most of the "work around the work" that fills a typical week (the reconciliation, the tracker, the deck nobody scheduled time for) doesn't fit into 20-minute windows between meetings.
By decoupling Cowork from a single device, Anthropic is targeting exactly that gap. A scheduled task can now run overnight through email threads, meeting transcripts, and recent news, then leave a briefing document and a drafted-but-unsent follow-up email ready for review over coffee. Nothing ships automatically — when Claude reaches a decision only a human can make, it sends the question to your phone, and you can redirect a draft mid-meeting without breaking the task's progress.
This distinction — asking AI for answers versus handing it an entire job — is the shift Anthropic is betting on. As one early user, Armmand Hosseini from Ramp's Customer Success team, put it: he started a client-tracking dashboard on his laptop and picked up the same session on his phone while waiting for his bag at baggage claim. "It just held the thread," he said. That continuity is the whole point.
### A platform-wide consolidation
On web and desktop, chat and Cowork now share a single home screen, and projects and artifacts persist across both surfaces. Anthropic frames it simply: delegating work to Claude is now as easy as asking it a question.
How to use it today
Getting started doesn't require much setup. On the web, launch a Cowork session directly from the home screen at claude.ai. On mobile, open Cowork from the sidebar in the Claude app for iOS and Android. For the deepest integration — including access to local files and your browser — download the Claude desktop app, which remains the best environment for heavy, file-intensive work.
The practical advice from Anthropic is to start with something already sitting on your plate: point Claude at a folder, an email thread, or a half-finished deck, and describe what "done" looks like. From there, Claude works through the material and surfaces questions only when a genuine judgment call is needed.
If you're testing lightweight AI workflows before committing to a full Cowork setup, a free toolkit like [mykreatool.com](https://mykreatool.com) is a useful sandbox for experimenting with AI-assisted content and automation tasks without any setup overhead.
To mark the mobile and web launch, Anthropic is also doubling Cowork usage limits through August 5, 2026 — a window worth using if you're on a Max plan and want to stress-test the feature on real work.
### Where to check availability
Feature availability differs slightly by surface (web, mobile, desktop), so check Anthropic's Help Center before assuming full parity across devices — some capabilities, like local file and browser access, remain desktop-only for now.
Who benefits
The clearest winners are knowledge workers whose jobs involve recurring, multi-step admin tasks that don't fit into a single sitting: finance teams building variance reports, customer success managers preparing account reviews, sales teams assembling decks from scattered call notes, and operations staff turning unstructured documents into trackers. Anthropic's own data — business operations and content creation making up roughly half of usage — backs this up directly.
Remote and travel-heavy professionals also gain meaningfully, since sessions now persist across a laptop, a phone, and idle moments in between. And critically, users who previously couldn't install a desktop app — due to IT restrictions or device limitations — can now access Cowork through the browser alone, widening the pool of people who can actually use it.
Risks
Handing AI agents overnight autonomy over email drafts, calendars, and business documents raises real questions about oversight. Anthropic's safeguard is that Claude pauses at decision points only a human can make and leaves drafts unsent for review — but that design still depends on users actually reviewing output carefully rather than rubber-stamping it because it arrived polished.
There's also a data exposure consideration: connecting email, messaging apps, and file systems to an agent that runs unattended increases the surface area for mistakes, whether that's a misfiled document or a draft sent to the wrong recipient before review. Organizations adopting this at scale should set clear policies about what tasks are safe to schedule unsupervised versus what still requires a human in the loop before execution, not just before shipping.
Conclusion
Claude Cowork's expansion to mobile and web turns AI agents from a desktop novelty into a genuinely persistent collaborator — one that keeps working through the night, across meetings, and between devices. With over 90% of real-world usage centered on everyday business operations rather than coding, this update targets the unglamorous, time-consuming work that fills most people's weeks. The doubled usage limits through August 5, 2026 make now a practical time to test it, provided teams stay deliberate about what they let run unattended.
Comments 0