
Anthropic has equipped Claude with the capability to control a Mac on behalf of the user. Within Claude Cowork and Claude Code, this AI model is able to independently open documents, navigate web browsers, execute development tools, and accomplish objectives via the computer’s user interface. Users can delegate tasks to Claude directly from a smartphone using the Dispatch feature.
The company showcased this new functionality via a demonstration video on YouTube. In the recording, Claude operates using both Cowork and Code modes, autonomously handling assignments and mimicking user actions—ranging from opening and modifying files to completing highly complex operations. This feature is slated for release on macOS, accessible initially as a research preview to subscribers of the paid tiers, Claude Pro and Claude Max.
Essentially, Anthropic is now actively implementing the very concept that was previously associated with the OpenClaw AI agent. The development trajectory for AI agents initially gathered momentum around Claude, even preceding OpenAI’s involvement in this specific area. Currently, this field is seeing participation from numerous other entities, including Perplexity Computer and Meta’s Manus.