Anthropic Introduces Claude Computer-Control Feature for Pro and Max Subscribers
Key Points
- Claude can control a MacOS computer by typing, clicking, and scrolling.
- The feature works with connectors for apps like Google Calendar and Slack.
- Claude asks for user permission before each action and can be stopped at any time.
- Available as a research preview for Claude Pro and Claude Max subscribers only.
- Supported on MacOS computers; some sensitive apps are disabled by default.
- Integrated with Anthropic's Dispatch service for mobile task delegation.
- Safety safeguards include scanning for prompt‑injection attacks and other vulnerabilities.
- Anthropic advises against using the feature with apps that handle sensitive data.
Anthropic announced that its Claude AI can now control a MacOS computer, allowing it to perform tasks such as opening files, scrolling, clicking, and using apps like Google Calendar or Slack. The capability is limited to Claude Pro and Claude Max subscribers, requires permission before each action, and includes safety safeguards to block prompt injections and other vulnerabilities. Users are advised not to use the feature with apps that handle sensitive data. The new function works with Anthropic's Dispatch service, enabling task delegation from a phone and supporting morning briefings or test runs.
Claude Gains Direct Computer Interaction on MacOS
Anthropic revealed that its conversational AI, Claude, can now act as an agent on a MacOS computer. The feature lets Claude manipulate the desktop just as a human would—typing, moving the cursor, scrolling, and clicking—while integrating with popular tools through connectors, including Google Calendar, Slack, web browsers, development utilities, and open files.
To initiate a task, Claude first asks the user for permission. The user can cancel at any point, and the system continuously scans for prompt‑injection attacks and other security risks. Anthropic has built safeguards designed to minimize the chance that the AI could be hijacked or cause unintended actions.
Availability and Subscription Requirements
The computer‑use capability is offered as a research preview to subscribers of Claude Pro and Claude Max. It is currently supported only on computers running MacOS. Anthropic notes that some applications that handle sensitive data are disabled by default, and it recommends users avoid using the feature with those apps.
Integration with Dispatch
Claude’s new ability pairs with Anthropic’s Dispatch service, which lets users assign tasks to Claude via a mobile device. Together, they can automate routine activities such as checking email each morning, launching a Claude Cowork or Claude Code session, or generating a morning briefing. While the feature is still early in development and complex tasks may not succeed on the first try, Anthropic is collecting user feedback to refine the tool.
Safety Considerations
Experts have cautioned that giving an AI direct access to a computer could expose users to security threats if the system is compromised. Anthropic acknowledges these concerns, emphasizing that the feature includes automatic scanning for vulnerabilities and that users should be mindful of the apps they enable.
Overall, the announcement marks a significant step toward more agentic AI capabilities, allowing Claude to perform practical, hands‑on tasks while incorporating multiple layers of user consent and security monitoring.