GitHub Copilot app Runs Multiple AI Agents in Parallel

This feature addresses the bottleneck of managing multiple AI coding tasks sequentially. Previously, a developer might have to wait for one agent task to finish before starting the next, or risk them conflicting on the same code. By isolating each session on its own Git worktree, GitHub enables true parallelism.

The practical benefit is in the workflow. Instead of watching agents work, the developer's role shifts to reviewing completed results and making decisions. The user interface, with a sessions view showing the title and progress of each card, is designed to manage this concurrent workflow.

The example in the post—adding a funded sort feature, performing an accessibility review, and running tests—shows concurrent tasks of different types. The post frames the experience as less chaotic than it sounds, comparing agents to separate washing machines that run at their own settings.