Anthropic's Claude Can Now Control Your Mac

In Brief

• Claude can now directly operate users' computers for task completion.
• People can assign tasks from a phone via Dispatch.
• Claude asks permission before accessing each app and sensitive apps are blocked by default.

Anthropic's latest research preview hands Claude direct control over users' computers — browsers, mice, keyboards and screens — to complete tasks on a person's behalf. The capability is available to Claude Pro and Max subscribers within Claude Cowork and Claude Code on macOS.

Table of Contents

• How Claude Computer Use Works
• Anthropic Sets the Stage for IPO

How Claude Computer Use Works

Claude cycles through a hierarchy of access before entering computer use, prioritizing connectors to services like Slack or Google Calendar when available. When no connector exists, Claude moves to Chrome to navigate the web. If those two options are unavailable, Claude directly manipulates system controls — scrolling, clicking and exploring as needed — after requesting explicit user permission. Users can end the session at any point.

Unlike Claude Cowork, which operates in a virtual machine, Claude computer use is interacting within your actual desktop and apps. Its navigation is based on the system taking direct screenshots of your computer to understand navigation and layout.

The announcement included the following caveat: "It won’t always work perfectly: complex tasks sometimes need a second try, and working through your screen is slower than using a direct integration." While the company built safeguards including per-app permission, app blocklists and action review, where the system monitors for signs of prompt injection, Anthropic warns against using the tool in apps with access to healthcare, finances or personal records. 

Anthropic paired the release with Dispatch, a feature that enables users to assign Claude one-off or recurring tasks from mobile devices. Users can hand off work from their phones and retrieve completed results on their computers.

Anthropic Sets the Stage for IPO

Anthropic's pursuit of growth across consumer and enterprise markets is setting the stage for an expected 2026 public listing. The company secured $13 billion in funding in September 2025 at a $183 billion valuation, engaged IPO counsel in late 2025, and raised its 2026 revenue forecast by 20% to $18 billion in January 2026. It then closed a $30 billion Series G in February 2026 at a $380 billion valuation, cementing its position as one of the most valuable private companies in the world.

The enterprise market is the primary growth engine behind these numbers. Anthropic has been building Claude into the tools businesses already rely on, with Google opening Workspace to Claude in April 2025 and Asana integrating Claude in January 2026. The January 2026 launch of Cowork — a persistent, agentic workspace — followed by plugins in February, shows the company's ambition to make Claude a core part of how teams operate.

Its February 2026 acquisition of Vercept, focused on AI perception and interaction, is intended to further that ambition.

At the same time, Anthropic has continued investing on the consumer side as a foundation for that growth. The release of Claude Opus 4.6 in February 2026, with agentic coding and long-context retrieval capabilities, and Claude Sonnet 4.6 two weeks later, expanded what Pro and Max subscribers can do. Features like persistent memory, introduced in September 2025, suggest the company is using individual users as a testing ground for capabilities that will eventually roll out to enterprise customers.

Have a tip to share with our editorial team? Drop us a line: