Shut Those Laptops! Anthropic Puts Its Claude Cowork Agent on Your Phone

The era of needing to keep your laptop half-cracked open to ensure AI agents continue their work is officially over. Anthropic announced on Tuesday that Claude Cowork, its advanced AI agent designed to perform digital tasks autonomously, is expanding its reach beyond the confines of the desktop application. This significant update means users no longer need to maintain an active device session for Cowork to run scheduled tasks overnight, as limited versions of the agent are now accessible through Anthropic’s existing Claude smartphone app and web browser, eliminating the prior requirement for a desktop connection. This expansion dramatically enhances Cowork's utility. For instance, Anthropic’s launch video demonstrates a user requesting assistance with an impending business deal renewal. With a single prompt, Cowork efficiently gathers pertinent data from email threads, Slack channels, meeting transcripts, and recent online discussions. Subsequently, it leverages this information to generate a comprehensive reference document for the meeting and a pre-written email. Previously, such complex tasks necessitated an active desktop session; now, Cowork can continue operating and catching late-night incoming messages even after the user has clocked out, offering unprecedented flexibility. My initial experience with Claude Cowork in January was remarkably impressive. The agent consistently followed through on tasks, such as organizing a chaotic collection of screenshot files into neatly labeled folders and efficiently scheduling events on my calendar. While it wasn't flawless and presented potential security considerations like prompt injections, Cowork undeniably represented a significant leap in how everyday users could interact with their devices. This contrasts sharply with the previous "Dispatch" feature, which allowed mobile task requests but critically required the paired desktop computer to be awake and the app open. Anthropic's latest move aligns with a broader, accelerating trend in Silicon Valley towards always-running, semi-autonomous AI agents that users can control via simple text commands. This shift was notably sparked by the viral success of OpenClaw, a homebrew agent that gained traction in early 2026, with early adopters entrusting it with aspects of their online lives 24/7. Competitors quickly followed suit, with OpenAI launching its adaptive agent Codex and Google introducing Spark, their own always-on agent, while Anthropic focused on making its agents more user-friendly, evolving from developer-centric tools like Claude Code to general-user applications like Cowork. The revamped version of Cowork is initially rolling out as a beta to subscribers of Anthropic’s premium Max plan, priced at $100 per month, with plans for a subsequent rollout to the cheaper Pro tier ($20 a month). Its availability for free users remains uncertain. Alongside this launch, Anthropic released a report detailing Cowork's usage patterns, indicating a growing adoption among white-collar professionals. The report highlights "business process and operations" (e.g., data reports, checklists) and "content creation and copywriting" (e.g., slide decks, partnership proposals) as the two primary categories of recent usage. Both Anthropic and OpenAI are actively exploring ways to seamlessly integrate their popular chatbots and agents into a unified, smartphone-centric user experience. OpenAI introduced Codex Remote in June, allowing users to control desktop agents from their phones, and further enhanced its iOS app in July with extensive Codex task management capabilities. Anthropic's current release takes this integration a step further by merging the Claude chatbot interface directly with the Cowork agent for both browser and desktop versions, embodying a Silicon Valley vision where agentic automation becomes a fundamental aspect of how all users, not just developers, interact with their devices.