Meta-backed Manus brings AI agents to your laptop, steps into OpenClaw territory

Manus AI

• Manus launches desktop app for direct AI agent access on devices
• AI agent can now read, edit, and organize local files and apps
• AI actions need user approval to ensure privacy

Did our AI summary help?

Manus is taking a significant step forward in the AI agent race—moving from the cloud to your computer. The company has launched Manus Desktop, a new application that brings its general-purpose AI agent directly onto personal devices, unlocking deeper access and control.

From cloud to your computer

Until now, Manus operated largely through the cloud, with users interacting via a web interface. That model worked—but it also came with limits.

With the new desktop app, a feature called “My Computer” changes that dynamic. It allows the AI agent to interact directly with local files, apps, and system tools.

In practical terms, that means the agent can read, edit, and organise files, launch applications, and even control workflows on your machine.

What the AI agent can actually do

Manus says the feature is designed for complex, multi-step tasks. For example, users can ask the agent to organise thousands of images stored locally, or even build an app by working alongside coding tools installed on the system.

This builds on Manus’s existing integrations with services like Google Calendar and Gmail, extending its reach from the cloud into the device itself.

The move also puts Manus in more direct competition with OpenClaw, the fast-rising open-source project that runs locally on user devices. OpenClaw has quickly gained attention, with Jensen Huang calling it “the next ChatGPT.” Its open-source nature and local-first approach have helped spark a broader shift toward autonomous AI agents.

Manus, in contrast, remains a paid service—but is now adopting a similar direction by bringing agents closer to the user.

Giving AI this level of access raises obvious concerns. Letting an agent interact with files and applications directly introduces risks around privacy, security, and unintended actions—issues already being debated across the industry.

Manus says it’s addressing this with explicit user controls. Actions require approval, with options like “Allow Once” for one-off tasks or “Always Allow” for trusted workflows.

The launch also ties into Meta’s broader AI ambitions. Meta acquired Manus in late 2025 to strengthen its push into autonomous AI systems, with plans to integrate the technology across its ecosystem, including the Meta AI assistant.

At the same time, the deal has drawn scrutiny. Chinese regulators are reportedly reviewing the $2 billion acquisition over potential technology control concerns, though Meta has said it complied with all applicable laws.