Your AI Assistant Can Now Use Your Computer While You're Away
We talk a lot about AI automation: connecting tools, triggering workflows, saving time on repetitive tasks. But there's always been a gap. What about the stuff that doesn't have an API? The things that only live inside a browser tab or a desktop app, where no connector exists and no webhook fires?
That gap just got a lot smaller.
Anthropic recently launched computer use for Claude inside Claude Cowork and Claude Code. In short, Claude can now actually use your computer, moving the mouse, clicking buttons, opening files, navigating the browser, just like a person would. No setup. No plugins. It figures out what it needs to do and does it.
How it actually works
When you assign Claude a task, it doesn't just jump straight to controlling your screen. It tries the smart route first, using native connectors to tools like Slack or Google Calendar if they're available. Only when those aren't an option does it fall back to direct computer control.
That's a sensible approach. It means faster, more reliable results when integrations exist, and a capable fallback when they don't. From there, Claude can scroll, click, open applications, run tests in dev tools, whatever the task needs.
You stay in control the whole time. Claude will always ask for permission before accessing a new application, and you can stop it at any point. There's also built-in protection against prompt injection attacks, with the system automatically scanning for that kind of activity in the background.
The piece that makes this useful for automation: Dispatch
Here's where it gets interesting from an automation standpoint. Alongside computer use, Anthropic also shipped a feature called Dispatch, available in both Claude Cowork and Claude Code.
Dispatch lets you assign tasks to Claude from your phone and pick up the results on your desktop later. It's a continuous conversation across devices. You send the instruction while you're on the go, Claude gets to work, and by the time you sit down at your computer, it's done.
Think about what that unlocks for daily workflows. You could have Claude check your inbox every morning and prepare a briefing before you get to your desk. You could ask it to pull weekly metrics, draft a summary, and drop it somewhere you'll find it. You could kick off a code session from your phone, have Claude make the changes, run the tests, and open a pull request, all without touching your laptop.
This is the kind of "set it and forget it" automation that used to require serious engineering effort. Now it's a message from your phone.
Worth being honest about the limitations
Computer use is still early. It's slower than a direct integration, complex tasks sometimes need a second attempt, and Anthropic is clear that it won't always get things right first time. For now it only works on macOS, and you'll want to keep sensitive data out of the picture while the technology matures.
But that's exactly the right way to release something like this, openly, with appropriate caution, and with the expectation that it improves through real-world use.
The bigger picture
For anyone building or using AI-powered automations, this is a meaningful shift. The bottleneck has never really been the AI's reasoning. It's been getting the AI connected to the right surfaces. Computer use punches through a lot of those barriers without requiring any technical setup on your end.
Pair that with the ability to assign tasks from anywhere via Dispatch, and you've got something that starts to feel like having an assistant who's genuinely autonomous. Not just a tool you query, but one that can go off and handle things while your attention is elsewhere.
It's available now for Claude Pro and Max subscribers. If you want to explore how this fits into your automation stack, that's exactly what we're here to help with at NTS AI Systems.