Claude Cowork Now Works When Your Laptop Is Closed: What Cowork on Web and Mobile Means for Your Team
TL;DR
Anthropic just made Claude Cowork — its AI worker that actually does multi-step tasks for you, not just chats — available on the web and your phone, and moved the work off your laptop and onto Anthropic's servers. That second part is the real news. Until now, Cowork ran on your own computer: close the lid and the work stopped. Now Cowork runs remotely in an isolated environment tied to your account, so a task keeps going after you shut your laptop, you can start it at your desk and check on it from your phone on the train, and scheduled jobs run on their own with no device turned on at all. In plain terms, Claude went from “an assistant that only works while you're sitting there watching it” to “a coworker you can hand something to and walk away from.” Anthropic is rolling it out in beta starting with Max users and doubling Cowork usage limits through August 5 to mark the launch. Here's what actually changed, why “it runs without your laptop” is a bigger deal than it sounds, and what your non-technical team should do about it this week.
First, a Reminder: What Cowork Actually Is
Claude Cowork is the version of Claude that does the work, rather than telling you how to do it.
If you've only ever used Claude as a chat box — ask a question, get an answer, copy it out — Cowork is a different mode. You hand it a whole task: “Go through these forty support emails, group them by issue, and draft a reply to each.” Cowork reads your files, makes a plan, works through the steps, and comes back with finished output while you do something else. It's less “search engine you talk to” and more “junior colleague you delegate to.”
That distinction matters for what changed this week. A chat box doesn't need to keep running when you look away — you're there for every message. But a coworker doing a forty-minute job is different: the whole point is that you get to leave. And until now, you couldn't really leave, because Cowork lived on your laptop. That's the limitation Anthropic just removed.
What Actually Changed
Cowork moved from running on your computer to running on Anthropic's servers — and became reachable from the web and your phone, not just the desktop app.
Three concrete things shifted, and they're worth separating out because each solves a different everyday frustration.
1. The work now runs remotely, so closing your laptop doesn't stop it
Previously, Cowork did its work on your machine. If you closed the lid, put the computer to sleep, or lost your connection, the task paused or died. Now the work happens in a secure, isolated environment on Anthropic's servers, tied to your account. You kick off a task, close your laptop, get on with your day, and it keeps going. When you come back, the finished result is waiting.
The everyday version of this: you ask Cowork at 4:45pm to compile next week's board pack from a dozen documents, then you shut your laptop and go home. It finishes at 5:20 without you. Before, you'd have had to sit there watching a progress bar until it was done.
2. It's on your phone and in a browser, not locked to the desktop app
Cowork used to be a desktop-only app. Now the same sessions live on the web and on mobile. Because the work runs on Anthropic's servers and is saved to your account, you can start a task on your work computer, glance at how it's going from your phone during a meeting, nudge it (“actually, leave out the Q3 numbers”), and pick up the finished document later from whatever device you happen to have. The session follows you, not the machine.
3. Scheduled tasks now run with no device on at all
This is the sleeper feature. You can save a task and have Claude run it on a schedule — every Monday at 6am, say — and because it's running on Anthropic's servers, nothing of yours needs to be turned on. Your laptop can be shut in a bag. Anthropic's example: set “Monday client prep” for 6am, and Claude works through your email threads, call transcripts, and recent news to build a briefing document that's sitting ready when you open your laptop with your coffee.
Previously a scheduled task still needed your computer awake and online. That's a meaningful difference: a scheduled job that requires your laptop to be on at 6am is a job that quietly fails half the time. One that runs on a server actually happens.
Why “It Runs Without Your Laptop” Is the Whole Point
The gap between an AI that needs you present and one that doesn't is the gap between a tool and a teammate.
Think about what makes delegation valuable with a human colleague. It isn't just that they're competent — it's that once you've handed something off, you get your attention back. You don't stand over their shoulder for the next hour. You go do the next thing, and they surface when it's done. That “you get to leave” property is most of what makes delegation worth doing at all.
AI that runs on your own device never fully had that property. Technically you could start a task and look away, but you were tethered — you couldn't close the laptop, couldn't leave for lunch, couldn't go into a meeting without risking the work stopping. So in practice most people babysat it. That's exhausting, and it caps how much you're willing to hand over.
Moving the work to a server cuts the tether. Now “hand it off and walk away” is literally true. And once that's true, the kinds of tasks you're willing to delegate get bigger, because the cost of delegating — sitting there watching — drops to zero. This is the quiet shift: not that Claude got smarter this week, but that using it stopped requiring your continuous presence.
A Plain-Language Analogy
The old Cowork was a power tool. The new one is a contractor with keys to the workshop.
A power drill is genuinely useful, but you have to be holding it. It only works while your hand is on the trigger. Put it down and nothing happens. That was Cowork on your laptop — powerful, but only while you were actively running it.
A contractor you've given workshop access to is different. You describe the job, hand over the keys, and go home. They work overnight; you come back to a finished result. You can call to check in or change the plan, but you don't have to be there. That's Cowork now: the work lives somewhere that stays open whether or not you're present, and you drop in from any device to steer or collect.
The intelligence didn't change between those two pictures. What changed is whether the work depends on you standing there. That's the entire upgrade — and it's why it matters more than it first sounds.
What This Looks Like for a Non-Technical Team
The value shows up in the recurring, low-glamour work that eats your team's mornings and evenings — the stuff that's worth doing but never worth doing yourself.
A few realistic examples of tasks that get genuinely easier now that the work runs on a server and can be scheduled:
- The Monday briefing. Every Monday at 6am, Claude reads last week's email with a key client, the notes from your last call, and any relevant news, and drafts a one-page “where things stand” before you're even awake. You edit and send. No one on your team spent Sunday night on it.
- End-of-day cleanup. You finish a run of customer calls, hand Cowork the transcripts at 5pm, close your laptop, and go home. By the time you're on the couch, it's logged each call, drafted the follow-ups, and flagged the two that need a real decision from you.
- The report nobody wants to build. Kick off the monthly numbers pull from twelve spreadsheets, then leave for a meeting. It compiles while you're gone. You review the draft after, instead of babysitting the assembly.
- Steering from your phone. A big task is running back at the office. From your phone between meetings you check the progress, notice it's about to include something outdated, and type one line to redirect it — without going back to your desk.
None of these are exotic. They're the ordinary preparation-and-follow-up work that surrounds real work, the kind that's easy to skip when you're busy and expensive to skip when it matters. Being able to schedule it, or hand it off and leave, is exactly what makes it finally get done.
The Honest Caveats
Two things to keep straight before you get excited: it's an early beta, and “runs on Anthropic's servers” is worth understanding, not fearing.
First, this is rolling out in beta, starting with Max-plan users and expanding to more plans over the coming weeks. If you're not on Max yet, you may not see it immediately. Treat the next few weeks as “try it on tasks you can afford to double-check,” not “wire it into anything critical unattended.” That's the right posture for any beta.
Second, the phrase “the work runs on Anthropic's servers” sometimes makes cautious teams nervous — where's my data going? The reasonable framing: this is the same kind of setup you already trust for your email, your files, and your CRM, which also live on someone else's servers. Anthropic runs each Cowork session in an isolated environment tied to your account. As with any tool that touches sensitive data, the sensible move is to start on lower-stakes tasks, understand what you're connecting it to, and expand as your confidence grows — the same way you'd onboard a new hire before handing them the crown-jewel accounts.
What Your Team Should Do This Week
You don't need a rollout plan. You need one recurring task and one hand-off-and-leave task to feel the difference.
1. Pick one thing that happens every week and schedule it
Find a small, repetitive prep task someone does every Monday — a status summary, a lead list, a “what changed since Friday” note — and set it up as a scheduled Cowork task that runs before the workday starts. The test is simple: does a useful draft appear on its own, without anyone having to remember to make it? That's the “runs with no device on” feature earning its keep.
2. Deliberately hand off a task and close your laptop
Give Cowork something that takes twenty or thirty minutes — compiling a document, working through a batch of emails — then close your laptop and walk away on purpose. The point is to feel that you're allowed to leave now. Come back and check the result. That single experience reframes what you think AI is for.
3. Try steering a task from your phone
Once during the week, start a task on your computer and then open it on your phone to check in or adjust it. Not because you'll do this constantly, but because it teaches you that the work follows your account, not your machine — which is the mental model that unlocks everything else.
4. Use the doubled limits window to experiment
Anthropic doubled Cowork usage limits through August 5 to mark the launch. That's a free window to be generous with experiments — hand off more, schedule more, try tasks you'd normally ration. Use it to find the two or three recurring jobs that are genuinely worth automating, before the limits return to normal.
FAQ
What is Claude Cowork in one sentence?
It's the mode of Claude that carries out whole multi-step tasks for you — reading your files, making a plan, and delivering finished work — rather than just answering questions in a chat, and it now runs on Anthropic's servers so it works across your desktop, the web, and your phone.
What's actually new here — didn't Cowork already exist?
Cowork existed, but it ran on your own computer and was desktop-only, so the work stopped when you closed your laptop. The new part is that the work now runs remotely on Anthropic's servers and is reachable from the web and mobile. That means tasks keep going after you close your laptop, follow you across devices, and — for scheduled jobs — run with no device of yours turned on at all.
How is this different from the “scheduled agents” you wrote about before?
Those are related but separate. The earlier scheduled-agents feature was aimed at Claude's managed agents — more of a behind-the-scenes, developer-flavored automation. This is scheduling inside Cowork, the everyday assistant a non-technical person uses directly. The headline improvement is that Cowork's scheduled tasks now run entirely on Anthropic's servers, so nothing of yours has to be awake for them to happen.
Do I need to keep my computer on for a scheduled task to run?
No — that's the point. Because the work runs on Anthropic's servers, a scheduled Cowork task runs even if your laptop is shut in a bag and every device you own is off. You just find the finished result waiting when you next open the app.
Is it safe to have AI running tasks on a server when I'm not watching?
Each Cowork session runs in an isolated environment tied to your account, the same broad model of trust you already use for cloud email, cloud files, and your CRM. The sensible approach with any new capability is to start on lower-stakes tasks you can double-check, understand what you've connected it to, and expand as your confidence grows. It's in beta, so treat the first few weeks as supervised experimentation, not unattended production.
Who can use it right now, and what does it cost to try?
It's rolling out in beta starting with Max-plan users and expanding to more plans over the following weeks, so availability depends on your plan and timing. To mark the launch, Anthropic doubled Cowork usage limits through August 5, which makes this a good window to experiment more freely than usual.
What's the one thing to take away?
That Claude just crossed the line from “an assistant that only works while you're watching it” to “a coworker you can hand something to and walk away from.” The intelligence didn't change this week — your ability to leave the room did. Start with one scheduled Monday task and one hand-off-and-close-the-laptop task, and you'll feel the difference immediately.
Want help figuring out which of your team's recurring tasks are worth handing to Cowork — and setting them up so they run reliably without anyone babysitting them? The Deployed Kickstart gets your team hands-on with Claude in a single day, mapped to your real workflows. The Partner program gives you ongoing support to roll it out across the business.