Claude Cowork Is on Your Phone Now: What That Actually Unlocks

Cowork has been my favorite Anthropic surface since it shipped, and it's also had one annoying constraint I kept running into: everything it did lived on whichever Mac it was running on. Set up a scheduled task, a watchdog, a daily brief, and the honest answer to "will this run tonight" was "only if your laptop is awake and online." I built a fair amount of automation around that limitation. Turns out I didn't have to.

On Tuesday, July 7, Anthropic shipped Cowork on web and mobile, and buried in the announcement is the part that actually matters: Cowork can now run in the background with no device online at all. Not "the app is on your phone now, cool." The task itself no longer needs a machine tethered to it.

What shipped

Cowork launched as a desktop app back in January. As of this week it's rolling out to claude.ai in the browser and to the sidebar of the Claude app on iOS and Android. It's in beta, starting with Max subscribers, with Pro, Team, and Enterprise access following over the next several weeks (per TechCrunch's and 9to5Mac's coverage of the launch).

The three things that actually change your workflow:

Sessions and files follow you across devices. Start a task at your desk, glance at progress from your phone in a meeting, come back and pick up the finished output on whichever screen is in front of you. It's the same session, not three disconnected apps pretending to sync.

Work continues after you close the laptop. This is the one I care about. Scheduled tasks previously needed a stable connection on the machine that owned them. Now they run headless, in the cloud, regardless of whether any of your devices are on.

The decisions still come to you. When Claude hits a call only a human can make, it doesn't guess and it doesn't stall silently. It sends a notification, you answer from whatever's in your pocket, and it picks the work back up.

Why this is the good fix, not a nice-to-have

I've built scheduled Cowork jobs before: a morning brief that pulls from a few connected tools and drafts a summary, a watchdog that checks for something going wrong and only speaks up when it finds it. Both are the kind of automation a solo operator actually wants, run it once, let it work while you sleep. The gap was always the same: "run while you sleep" quietly meant "run while your laptop happens to be open and connected," which for anyone who actually closes their laptop at night was never true.

That's the constraint this release removes. It's not a UI convenience, it's the difference between an automation you can trust and one you have to babysit.

It also lines up with how people are actually using Cowork, which surprised me a little. Anthropic said that across roughly 1.2 million Cowork sessions, more than 90% of the work wasn't software development at all. The largest chunk was business operations work: reconciling spreadsheets, turning scattered updates into a report, building a deck from call transcripts. Content and copywriting was next. Code was a small slice.

If you've been thinking about Cowork as "the agent that helps engineers," that's not what most people are actually pointing it at. It's closer to an ops hire that never sleeps, and this release is Anthropic finally matching the product to that reality: an ops hire needs to work independent of whether you happen to have your laptop open.

What I'd actually do with it this week

If you already have a Cowork task set up, check whether it's one that only worked because your machine stayed on. Those are the ones that get more reliable overnight, literally, with zero changes on your end once the rollout reaches your plan.

If you haven't built anything in Cowork yet, this is a better starting point than it was two weeks ago. The barrier used to be "I need to remember to leave my laptop running," which quietly kills a lot of good automation ideas before they're built. That barrier is gone. Start with something boring and real: a weekly report you currently build by hand, an inbox you'd like triaged before you open it, a metrics check you'd like to run daily without thinking about it.

Anthropic also doubled Cowork's five-hour usage limits through August 5 for eligible Pro, Max, Team, and legacy seat-based Enterprise users, so if you've been rationing runs, this is the window to actually stress-test something bigger.

The honest caveats

It's beta, and it's starting with Max subscribers only, with Pro, Team, and Enterprise following over the coming weeks, so most people reading this don't have it in hand yet. Web and mobile access is about reach and continuity, not new capability. Cowork isn't doing anything on your phone that it couldn't already do on desktop, you're just no longer the single point of failure for whether it keeps running.

That's still the right thing to ship first. An agent that can act across your calendar, your files, and your inbox should ask before it makes a call only you can make, and the notification loop is how that stays true even when you're not sitting at the machine. Reach without that check would be a worse product, not a better one.

Worth trying the moment it lands on your plan. If you want a walkthrough of setting up your first real Cowork automation, that's exactly the kind of thing we build live in the 2-Day Intensive.