Anthropic has added a new feature to the Claude mobile app that lets you check on work your AI assistant is doing on your computer—right from your phone. You can see what's happening and approve actions when Claude needs permission, like sending an email or saving a file. However, you still can't start new tasks from your phone; Claude has to be running on your desktop to actually do the work Engadget.

Here's what made this phone feature possible: Anthropic changed how Claude handles long-running tasks on your computer. Before, if your internet connection dropped while Claude was working on something, the task would stop or fail. Now Claude can keep working even if your Wi-Fi cuts out or your laptop goes to sleep. That's the real reason the phone notifications are actually useful—without that change in how tasks run, checking your phone wouldn't matter much.

Before Claude takes actions that affect your computer, it asks for your permission. You'll get a notification on your phone, and you can approve or reject it without sitting at your desk. Anthropic says: "Nothing ships until you've reviewed and approved it" Engadget. This approval step keeps Claude from doing too much on its own.

The phone feature is starting with Max subscribers, which is Anthropic's most expensive plan. The company plans to expand it to other plans over the coming weeks, though no exact date was announced.

Anthropic also said it plans to combine this task feature with regular Claude chat into one unified interface, beginning on the web and desktop apps. The timeline for that hasn't been set.

Why the background update matters more than the phone feature

The phone notifications look like the big news, but the more important change is how tasks can now run in the background without needing a constant internet connection. Think of it like a timer running on your oven: if you close the door, the timer keeps working. Before this update, Claude's tasks were more like a video call—if the connection dropped, everything stopped.

The approval system also deserves attention. AI assistants that can do more work on their own are more useful, but people also want to stay in control. Having Claude ask your phone for permission before it sends an email or changes a file is a practical middle ground: Claude can keep working, but you're always in the loop on the important stuff. The company didn't say exactly which actions are auto-approved and which ones ask for permission, so that's worth checking if you use this.

Longer term, Anthropic is planning to merge this task-running feature with the regular Claude chat into one app. Right now, they're separate tools, but the company seems to be deciding they're really just different ways to use the same AI—more like switching modes in a single app rather than using two different tools. Other AI companies like Microsoft are doing something similar.

For now, only the most expensive plan gets the phone feature, but that follows a pattern where companies test new features with their premium customers before rolling them out to everyone. How valuable this phone feature turns out to be will likely depend on who's using it—it probably matters more to people running complex batch jobs than to casual users checking email.