What is the simplest way to let AI control your computer?
Install a desktop app that already has computer use built in, turn the feature on, approve the permission prompt your system shows, and type what you want done. That is the whole route for most people. You do not need a virtual machine, an API key or a line of code, which is what the developer documentation for these models assumes.
Two routes, and which one is yours
The developer route starts from a model's API. Anthropic's computer use tool, for example, gives a model screenshot, mouse and keyboard control, but your own program has to run every action in an environment you build, and the docs advise a dedicated virtual machine or container. That is the right road if you are making a product. It is a long one if you only want your expenses typed into a spreadsheet.
The app route hides all of that. Claude's desktop app and superbot both ship computer use as a setting, and both run on macOS and Windows. You install them like any other program.
Five steps that work for most people
- Pick one app and install it from the maker's own site.
- Find the computer use switch. In superbot it sits in the message box and moves through Off, Ask and Always; it starts at Off. In Claude it is in the desktop app's settings.
- On a Mac, say yes when the app asks for Accessibility and Screen Recording. On Windows superbot shows an Allow card in the chat instead.
- Ask for one small thing in a sentence, such as "make a folder called Receipts on my desktop and move every PDF from Downloads into it".
- Watch the first few runs before you let it act unattended.
What simple does not mean
Simple to start is not the same as quick to finish. Anthropic notes that working through a screen is slower than using a direct connection to a service, and that complex jobs sometimes need a second attempt. Pick tasks where a slow run still saves you time, and where a mistake is easy to spot.
Trying before you install
If you would rather see an agent work before giving it your computer, superbot in the browser needs no account. There it can open web pages in a browser of its own to read them, click and fill in forms; using your own screen needs the desktop app.
Sources
Last updated .