Claude Cowork goes mobile and Anthropic has new insights into its usage

Anthropic is rolling out its AI agent Claude Cowork to mobile devices and the web, the company announced in an official blog post. The tool, previously limited to the desktop app, now lets users start a task on a laptop, check progress on a phone, and retrieve finished work from any browser. Beta access begins with Max subscribers and will expand to other plans in the coming weeks.

Cowork differs from a chatbot. Users hand it a task, and it works across files, calendar, email, messaging apps, and connected tools until the job is finished. With the update, tasks can also run without any device switched on. Anthropic gives an example: schedule a client briefing for 6 a.m., and Claude works through email threads and transcripts overnight, drafting a briefing document and an unsent follow-up email ready for review over coffee.

When Claude reaches a decision only a human can make, it sends a question to the user’s phone. “Nothing ships until you’ve reviewed and approved it,” the company states. Desktop remains the fullest version of Cowork, with access to local files and the browser. To mark the launch, Anthropic extends doubled usage limits through August 5.

Most sessions are not about code

Alongside the mobile launch, Anthropic publishes an analysis of 1.2 million anonymized Cowork sessions collected between May 11 and May 31, spanning more than 600,000 organizations. The results challenge assumptions about AI agents built on coding tools like Claude Code. More than 90 percent of sampled sessions had nothing to do with software development, as David Gewirtz reports for ZDNET.

Business process and operations made up the largest share at 33.4 percent, covering tasks like reconciling spreadsheets or building status reports. Content creation and copywriting followed at 16.4 percent. Together, the two categories account for roughly half of all usage. Software development trailed at just 8.7 percent.

Anthropic calls this pattern “the work around the work”: tasks that rarely appear in a job description but consume a large share of everyone’s week. Michael Nuñez, writing for VentureBeat, notes that Anthropic frames this as a deliberate strategy: rather than replacing expert judgment, Cowork handles the administrative tasks surrounding it, from a lawyer’s document formatting to a hiring manager’s interview notes.

Context: growth, competition and scrutiny

The launch lands during a busy stretch for Anthropic. The company recently released Sonnet 5, the model now powering Cowork, and Claude Tag, a Slack-based collaboration agent. Nuñez points out that Anthropic increasingly positions itself as an enterprise-wide platform, citing a Ramp AI Index showing Anthropic ahead of OpenAI in business adoption for the first time.

That growth comes with friction. Gewirtz recalls a recently disclosed sandbox escape vulnerability in Cowork’s desktop version, which Anthropic said required existing local access to exploit. Moving tasks to the cloud changes that risk profile, though it raises separate questions about data handling for autonomous, unsupervised tasks. Nuñez also notes that Alibaba has banned Anthropic tools for employees amid a dispute over alleged model distillation, complicating the company’s global enterprise ambitions even as it invests billions in new data center capacity.

Sources

• Claude Cowork on web and mobile: hand off work anywhere – Claude by Anthropic
• Anthropic’s Claude Cowork heads to the cloud as data shows 90% of sessions aren’t for coding – ZDNET
• Anthropic brings Claude Cowork to mobile and web as usage data shows most users aren’t coding – VentureBeat