Anthropic has rolled out a significant update to its Claude AI tools, enabling Claude Code and Claude Cowork to interact directly with users’ computers on macOS. This enhancement allows these AI agents to open files, navigate web browsers, and run developer tools, marking a notable step toward more autonomous AI assistants on macOS devices.
Initially, Claude will attempt to complete tasks using pre-integrated services like Google Workspace or Slack. However, when these connectors aren’t available, Claude can perform actions on the local machine after requesting user permission. Despite these safeguards, Anthropic advises against deploying this functionality with sensitive data to avoid potential security risks.
Availability and use cases for Claude AI computer control on macOS
The new computer control features are currently accessible only to Claude Pro and Claude Max subscribers using macOS. This rollout is part of an ongoing research preview, with Anthropic actively seeking feedback to refine the experience. The integration also works alongside Anthropic’s Dispatch platform, which enables continuous conversations with Claude across devices, including phones and desktops.
Claude Cowork, introduced earlier this year as a more user-friendly version of Claude Code aimed at programmers, benefits particularly from these enhancements. The ability to autonomously manage files and tools will likely expand workflows for developers and casual users alike by reducing manual overhead.
This direct interaction with the computer places Anthropic ahead in the race among AI providers to offer more proactive and efficient AI assistants. However, the cautious rollout and upfront warnings highlight the delicate balance companies must strike between enabling automation and safeguarding user privacy and security.