Anthropic’s Claude Code on the web is picking up a practical new capability: scheduled tasks that run in the cloud, even after the browser tab is closed. The update lands shortly after the recently announced /loop feature, and it pushes Claude Code’s web experience a little closer to something that behaves like a long-running automation surface rather than a session that needs constant supervision.
Scheduled tasks, without the tab babysitting
In her post, Lydia Hallie notes that “Claude Code on web now has scheduled tasks that run in the cloud,” adding that it’s now possible to close the tab and still have the task continue running.
That may sound like a small UX adjustment, but it changes the working model for AI-assisted coding on the web: instead of keeping an active session open, scheduled work can run independently and be checked on later.
Community reactions quickly converged on that theme. Martin S. summarized the appeal as finally being able to “stop babysitting a tab like it’s 2009,” while Mingta Kaivo framed it as a shift from something continuously attended to something periodically monitored—an interaction pattern that tends to encourage longer-running workflows.
Early questions: self-scheduling agents and control
Alongside excitement, the replies also surfaced questions about where this feature could lead.
A couple of developers asked about a more advanced pattern: creating scheduled tasks from within a scheduled task. Users pointed to the design implications—arguing that self-scheduling agents are qualitatively different from simple recurring jobs, and closer to “actual autonomous agents.”
Other replies raised broader concerns about who controls scheduled processes and the infrastructure behind “cloud tasks,” reflecting the natural tension that comes with moving more execution into managed environments.
The creative-use-case phase begins
With the mechanics now in place, some of the most interesting discussion is less about maintenance-style cron work and more about what people will build on top. Trevor I. Lasn called out that the “creative use cases” may end up being more compelling than the original intent—often the sign that a platform feature is about to get stress-tested in the wild.
