Claude Cowork arrives on mobile and web as agents move beyond coding

Claude Cowork now runs from your phone.

On July 7, 2026, Anthropic launched Claude Cowork on mobile and web as a beta, rolling out first to Max subscribers. The pitch is continuity: kick off a task from your laptop, steer it from your phone on the go, and collect the result wherever you are. Tasks keep running in the background even when your laptop is closed.

An agent that does not need you watching

The background execution is the meaningful part. Most AI tools still assume you are sitting in front of them, prompt by prompt. Cowork’s model is closer to delegation: hand off a task, walk away, check in from whatever device is nearby. Closing your laptop no longer kills the work.

Mobile becomes the steering wheel rather than the workbench. You are not doing the task on your phone; you are supervising an agent that is doing it elsewhere.

Who actually uses Cowork

The launch announcement came with a number that reframes the whole agent story: more than 90% of Cowork usage is not software development.

The breakdown, from Anthropic’s own data, puts business operations in the lead at 33.4%, with content creation and copywriting following at 16.4%. Coding, the use case that defined AI agents in the first place, is a small slice of what people actually hand off.

The agent era leaves the terminal

That usage split is the bigger story. Agents were built by developers, for developers, and the assumption was that code would stay their home turf. Instead, the people quietly adopting them at scale are running operations, writing copy, and managing the unglamorous middle of office work.

The agent era is moving from coders to everyone else, and the mobile launch accelerates that: the people most likely to steer an agent from a phone between meetings are not the ones living in a terminal. What would you hand off to an agent first? For a third of Cowork users, the answer is already business operations.