Claude Upgrades Its Computer Agent: Run PC Tasks 24/7 From Your Phone

According to 36Kr, Anthropic’s large language model Claude has recently demonstrated stronger “computer agent” capabilities: with just a single natural-language instruction on a phone, users can—once authorization is granted—have the AI take over and operate a personal computer, continuously completing a series of connected tasks and enabling an automated workflow that runs close to 24/7 without interruption.

From the industry’s perspective, this kind of “AI doing operations on your behalf” is seen as an important step toward a general-purpose intelligent assistant: it extends model capability from “generating content” to “taking actions,” and could be applied to everyday office work, information retrieval, and repetitive process automation. At the same time, social platforms have also seen discussions about Claude improving work efficiency, but most of these experiences are personal shares and still need validation with real business data.

Commentary: Once computer-level agents become widespread, the key question won’t be “what can it do,” but “what is it allowed to do.” Future product rollout will depend more on mechanisms such as least-privilege permissions, traceable logs, security audits, and safeguards against mis-operations, and enterprises will also need to update internal IT and compliance processes accordingly.