I have very mixed feelings about using Manus.
On one hand, I’m genuinely impressed by its high efficiency and task completion — it can quickly and accurately build product prototypes based on my needs.
On the other hand, when I invited my friends to try it, the system falsely flagged me for “fake referrals” and suspended my account — even though my friends actually subscribed.
I’ve reached out to their support team multiple times, but the issue has never been resolved.
I honestly don’t understand what their support team is doing. No matter how great a product is, poor customer service can easily ruin it.
Manus just took a big step forward with My Computer, bringing its AI agent out of the cloud and directly onto your desktop.
Until now, Manus worked in a cloud sandbox. But most of our real work lives locally: files, dev environments, apps, and workflows. My Computer bridges that gap by letting Manus execute command line instructions on your computer to read, organize, edit files, and control local applications.
What makes it interesting is the automation potential. Manus can organize messy folders, rename hundreds of files, build apps through CLI tools like Python, Node.js or Swift, and even run tasks using your machine’s idle compute.
You can also assign tasks remotely, for example, asking Manus to find a file on your home computer and email it through Gmail while you're away.
Key highlights:
Works directly with local files, tools and apps
Executes terminal commands with your approval
Automates repetitive file and workflow tasks
Can build software projects via CLI tools
Uses idle compute resources in the background
Lets you trigger tasks remotely across devices
This seems especially useful for developers, builders, and anyone managing large local workflows who wants automation beyond browser-based AI tools.
It reminds me of what Perplexity is doing with Perplexity Computer, but focused on letting an AI agent directly interact with your own machine and workflows.
What use cases are you thinking with My Computer by @Manus?
Moving Manus from cloud-only sandboxing to direct local machine access via CLI execution is the natural next step — most real productivity workflows involve local files, dev environments, and desktop apps that a cloud-only agent simply can't touch, so bridging that gap unlocks an entirely different class of automation tasks. The remote task triggering is a compelling feature for power users, but how does Manus handle permission scoping on local execution — is there a granular approval system for different command types, or does the user approve each terminal command individually?
so it's a claude desktop clone?