AI company Anthropic announced Tuesday that it is moving its popular Claude Cowork agent to the cloud, making it accessible to users from multiple devices, and able to perform tasks even when the devices are not online.

Subscribe to read this story ad-free

Get unlimited access to ad-free articles and exclusive content.

Cowork is a feature that was built primarily for nontechnical users. Launched in January, it automates multistep workflows for personal organization and scheduling by pulling from users’ files, folders and applications without the need for constant prompting or direction.

Up to now, Cowork had only been available for users operating on specific devices, which needed to remain open and active for work to be completed, and users could not operate Cowork via multiple devices using one account.

The update announced Tuesday will change that.

The changes announced will allow scheduled tasks, like drafting an email, to run without a device online, but Claude will still ask a user for a final review and approval before shipping, the company said in its blog.

Anthropic will give beta access to users with Max subscriptions, the company’s most premium plan for consumers, and will expand access to more plans in the coming weeks. To mark the launch, Anthropic is extending Cowork’s doubled usage limits through Aug. 5.

In addition to moving Cowork to the cloud, Anthropic says it will unify Claude chat and Cowork, putting them in the same space and giving them access to the same files.

Cowork competes in a crowded push toward “agentic” AI — systems that complete multistep tasks rather than answer questions — where Google, OpenAI and Microsoft have all launched rival agent products. Claude has become a favorite among technical and nontechnical users, though, even as OpenAI’s ChatGPT remains one of the most well known and most used AI products.

As agents have become increasingly popular, users have expressed concerns about the broad permissions that often need to be granted to them in order to be effective. Anthropic has attempted to address this via its permissions structure, which frequently and repeatedly asks for consent to perform specific tasks as Cowork operates.