Anthropic announced this Monday that the Claude chatbot can now perform tasks directly on the user’s computer remotely. The new feature integrates Claude Code and Claude Cowork, allowing commands sent by the browser or mobile app to trigger actions on the desktop. Antes, instructions needed to be given locally on the computer itself.
Users can delegate activities such as exporting files, attaching documents to meeting invitations, or browsing applications while performing other tasks in parallel. The system prioritizes dedicated connectors for direct integration with services, but uses screen, mouse and keyboard control when necessary. Activation occurs after explicit permission from the user.
- The feature works through Dispatch, which allows you to assign tasks from anywhere.
- Claude searches for the most efficient method to complete the submitted request.
- The execution takes place in the background on the computer, which must remain turned on.
How the Claude remote control works in practice
The assistant analyzes the requested task and decides whether to use native integrations or take visual control of the desktop. Ele can open programs, scroll pages, click buttons and enter information depending on the context of the request. Essa approach simulates actions that a human user would perform manually.
In the example disclosed, a professional late for a meeting asks Claude to export a presentation as a PDF and attach it to the corresponding invitation. The system locates the file, performs the conversion, opens the calendar application, and completes the attachment without further intervention. The interaction occurs even when the user accesses Claude via cell phone.
Protection mechanisms implemented by the company
Anthropic has built-in layers of security to reduce the risk of misuse or malicious commands. The system performs automatic checks on model activations to identify suspicious prompt injection patterns. Usuários can interrupt execution at any time during the process.
The company issues clear alerts when the feature is activated. Algumas actions performed by Claude may be irreversible, depending on the application involved. Authorized Aplicativos can serve as a bridge to access other programs installed on the computer.
- Documents or websites with potentially malicious instructions require special attention.
- Sensitive content on the screen must be closed before granting permission.
- Screen control allows Claude to view currently visible information.
Current availability and functionality limitations
The tool is in the testing phase as a research preview and presents variable performance in more complex tasks. Algumas operations may require multiple attempts or progress more slowly when they depend on manual control of the interface. Dedicated Conectores still offer greater speed and accuracy when available.
The functionality is currently restricted to macOS. Assinantes of plans Claude Pro and Claude Max have immediate access to the new feature. Claude Cowork serves users who seek automation in general knowledge activities, while Claude Code focuses on more technical development flows.
Details about required subscription plans
The Pro plan offers expanded usage for everyday tasks and includes access to new remote control capability. Já the Max plan delivers significantly higher processing limits, ideal for those who carry out intensive or simultaneous operations. Ambas options require active subscription to enable full feature.
Anthropic recommends that the computer remains turned on and connected to the internet for remote tasks to be completed successfully. Notificações indicate the progress and completion of actions performed by Claude. The Dispatch makes it easy to send commands directly from your smartphone, increasing flexibility of use.
Integration with existing Claude ecosystem tools
The new remote control complements previous features such as connectors, which allow direct interaction with external services without the need for screen navigation. Quando these paths are not available, the system activates the view and action mode on the desktop. Essa combination seeks to balance efficiency and range of possible applications.
Developers and knowledge workers can delegate repetitive or time-sensitive routines while focusing on higher-value activities. Running in the background keeps the computer available for other functions at the same time. The feature reinforces the trend of autonomous agents that act independently in the user’s local environment.
Safety Considerations and Responsible Use
Experts at Anthropic highlight the importance of granting permissions only after evaluating the context of the task. Internal activation scanning helps mitigate exploit attempts, but does not completely eliminate all potential risks. Usuários must maintain updated software and monitor activities when Claude is operating.
The functionality advances the concept of AI agents that go beyond conversational responses. Ela transforms Claude into a tool capable of physically interacting with the user’s computing environment. The initial focus on macOS reflects a controlled launch strategy before expansions to other platforms.