On Tuesday, Anthropic announced that Claude Cowork, its AI agent designed to perform digital tasks, is expanding beyond the desktop app. Users can now interact with limited versions of Cowork via the existing Claude smartphone app or web browser, without the formerly required desktop connection. The agent can run scheduled tasks overnight without leaving a device active. In a launch video, a user asks Cowork to pull together data from email threads, Slack channels, meeting transcripts and recent online chatter to generate a reference document and a prewritten email for a business deal renewal. Previously, users could pair their smartphone app with their desktop through the Dispatch feature, but Anthropic’s description noted, “Your computer must be awake, and the app must be open for Claude to work on tasks.” Now, Cowork can run tasks without an active desktop session. This announcement is part of a shift in Silicon Valley towards always-running, semi-autonomous AI agents controlled via texting, sparked by OpenClaw, a home-brew agent with a lobster mascot that went viral at the beginning of 2026. In the first half of the year, OpenAI launched Codex, Google launched Spark, and Anthropic leaned further into making its agents more user-friendly.