I Stopped Waiting Until I Got Home to Keep Working: What Claude Cowork on Mobile Actually Changed

Anthropic rolled its Claude Cowork tool out to mobile and web in July 2026, which means a task you hand to your AI assistant at your desk no longer has to sit there waiting for you to get back to it. After weeks of actually using that, here's what it's worth in practice.

The Problem I Actually Had

I use Claude for a lot of my work at try-ai.net, everything from building tools to drafting content to digging through research. The issue was never Claude itself. It was me. I'd hand it something substantial to work through, then remember I needed to run to the hardware store or pick up a prescription. Before this update, that meant the work just sat there. My laptop stayed open on the kitchen table like it was guarding the task, and I felt weirdly stuck near a machine for no good reason. That's the specific thing that changed. I don't have to choose between leaving the house and letting a task keep moving.

What Actually Shipped

Anthropic's July 2026 rollout wasn't just "now it works on your phone." Three pieces matter here, and they matter in this order:

Cross-device sync. You start a task on one device and pick up monitoring it on another, no re-launching, no re-explaining what you were doing.

Background execution. The work keeps running even if your device goes offline. Anthropic's own framing was that "your work goes everywhere with you, and keeps going without you," and that's a fair description of what it feels like day to day.

Human-in-the-loop approvals. When the AI hits a decision that actually needs a person, it surfaces that to your phone and waits. It doesn't barrel ahead on something you'd want a say in, and it doesn't stall the whole task waiting for you to be back at a keyboard either.

How I Actually Use It

Here's the real workflow, not the marketing version. I start a task on my laptop, often something technical since that's a lot of what I build. If I need to step away, I don't shut anything down, I just leave. On my phone, I can check what's happened, approve a step it flagged, or just glance at progress. When I'm back at the laptop, I pick up that exact same session, not a new one I have to catch up. No copying context between devices, no "wait, where did I leave off."

The part that surprised me is how much this changes what I'm willing to start before running an errand. Previously I'd wait to kick off anything that might need a decision partway through, since I didn't want to be stuck babysitting it. Now I just start it and go.

This Isn't Really a Developer Story

It's tempting to file this under "nice thing for people who code," but Anthropic's own usage data says otherwise. Looking at over a million anonymized Cowork sessions across 600,000-plus organizations, business processes and operations made up the largest chunk of usage at 33.4%, with content creation and copywriting at 16.4%. Software development came in at 8.7%, well behind both. Anthropic calls this "the work around the work," the administrative and connective tasks that eat time in almost any job but aren't the job itself.

That tracks with what I'd expect. A task doesn't need to be code to benefit from not stalling out the second you close your laptop. Drafting follow-up emails, organizing research, pulling together a status update, any of that can run in the background exactly the same way my Claude Code sessions do.

Setting This Up Yourself

If you're on a Cowork plan, mobile and web access should already be available rather than something you have to hunt for in settings. A few things worth knowing going in:

• Start tasks the same way you normally would, from whichever device is in front of you. The continuity is automatic, not something you configure per task.

• Expect approval prompts for anything consequential. Treat those as a feature, not friction. That's the system correctly refusing to guess on your behalf.

• Background execution means a task can finish while your phone is in your pocket. Check back rather than assuming silence means nothing happened.

Mistakes to Avoid

• Assuming you need to babysit it. The whole point is that you don't. If you're still hovering over every step, you're not using the background execution part at all.

• Ignoring approval prompts for hours. They exist because a decision point actually needs you. Letting them pile up defeats the "keeps going without you" promise for exactly the steps that matter most.

• Starting something ambiguous and walking away. Vague instructions plus no supervision is a bad combination. Be specific about what you want before you hand it off, especially if you won't be around to redirect it early.

• Forgetting it's still worth reviewing the result. Trust but verify still applies. Background execution changes when you check the work, not whether you should.

Prompts Worth Trying

Kick off a task you can walk away from:

I need to step away for about an hour. Before I go, start [TASK], and if you hit a decision that needs my input, flag it clearly and wait rather than guessing. Otherwise keep going and have a summary ready for when I'm back.

Catching up from your phone:

Give me a quick status on the task I started earlier: what's done, what's in progress, and anything you need a decision on from me right now.

Handing off between devices:

I'm switching from my phone back to my laptop. Before I do, summarize exactly where things stand so I can pick this up without losing context.

Try any of these the next time you're about to leave your desk mid-task. The difference isn't dramatic on any single use, it's more that it removes a decision you were making without noticing you were making it: wait, or leave the work behind.