Anthropic Rebuilds Claude Code for Parallel Multithreaded Tasks

Summary

Anthropic has rebuilt Claude Code around a parallel collaboration mechanism for complex software projects. Users describe a project goal, and a coordinator divides it into tasks assigned to multiple independent threads, each running in its own cloud session. Users can monitor the main chat and individual threads in real time, including from mobile devices. Each thread can run tests and submit pull requests, while Claude maintains shared memory across threads and the project repository collects uploaded files and generated results. Cloud sessions are currently being tested with some Pro and Max subscribers; Team and Enterprise access is planned for a later stage, and local execution is still under development. Users can apply through a waitlist. The update moves Claude Code beyond a single AI coding assistant toward a multi-agent development platform. Anthropic had previously made Claude Code’s autonomous mode the default and said it outperformed human developers on some safety tasks, although the article provides no supporting measurements. As more work is delegated to autonomous parallel agents, the article identifies computing-resource use, user costs, and developer control over agent behavior as emerging concerns.