Manus launches "My Computer" app: AI organizes photos, remote control your home computer to send files, and more

IT Home reported on March 18 that Meta’s AI company Manus released an announcement on March 16 announcing the launch of a “My Computer” desktop application. It can run on Windows 11 PCs and on Macs powered by Apple Silicon chips, turning ordinary computers into powerful personal AI agents.

IT Home has included a demonstration video as follows:

In terms of interaction mechanisms, the “My Computer” desktop app primarily carries out its work by executing command-line instructions (CLI) in a computer terminal. It can accurately read, analyze, and edit local files, greatly simplifying tedious and repetitive tasks.

For example, when faced with a disorganized photo library, Manus can quickly scan and automatically create subfolders to categorize and organize by content. When dealing with hundreds of invoices, it only needs to run a few lines of terminal commands to complete standardized renaming within minutes.

In addition to basic file management, the desktop app can seamlessly invoke local toolchains such as Python, Node.js, and even Swift. Internal testing shows that Manus used only 20 minutes to build from scratch, debug, and package a Mac app with real-time translation and subtitle features using pure terminal commands—throughout the process without any need for a person to manually open Xcode to write code.

It can also wake up an idle GPU to train machine learning models, or convert an idle Mac mini into an always-on standby remote AI node.

The desktop app can further seamlessly coordinate cloud intelligence with local computing power. Even when users are away and don’t have their computer with them, they can send instructions to the Manus at home from their phone.

The system will remotely wake the local machine, precisely locate the target files, and automatically call bound third-party cloud services such as Gmail to send them to the customer.

On the security front, Manus has built a strict security governance framework. The system requires that every terminal command must receive the user’s explicit approval before it can be executed. Users can flexibly choose between “Always allow” or “Allow only once” based on their level of trust.