The new update lets Claude click, type and open apps like a real user with permission
Anthropic has introduced a new feature that allows its Claude AI to use your computer to complete tasks. The update lets the chatbot open files, use apps and perform actions like clicking and typing, similar to how a person would use a keyboard and mouse.
The feature is now available in research preview for Claude Pro and Max users. It currently works on macOS and will expand based on feedback.
This marks a major shift in how AI tools are used.
Instead of just answering questions, Claude can now take action. It can check your files, use your browser and connect with tools like Google Calendar or Slack to complete tasks. If a connector is not available, Claude can still perform the task manually by interacting with the system.
The system is designed to ask for permission before taking any action. Users can also stop tasks at any time, giving them control over what the AI is doing.
But there are risks.
Giving an AI access to your computer raises security concerns. Experts say these tools can act quickly and may take actions users do not expect. There is also a risk of misuse if systems are not properly secured.
Anthropic says it has added safeguards.
The system can scan for threats like prompt injection attacks and limit access to sensitive apps. Some features are disabled by default to reduce risk.
This update comes as more companies build “agent style” AI tools. These systems go beyond chat and try to complete tasks automatically. Similar tools have been growing quickly, especially after open frameworks like OpenClaw gained attention earlier this year.
Claude’s new feature also works with Dispatch. This allows users to assign tasks from their phone, such as checking emails or preparing updates, even when they are not at their computer.
The feature is still early. Anthropic says some tasks may not work perfectly yet, and the goal of this release is to improve the system based on real user feedback.
(Image credit: Anthropic)
