Claude Code has relaunched its Projects feature, giving users a single environment in which to run multiple AI agents in the cloud. The revamped tool allows several agents to operate under the same roof, sharing a common memory, set of goals, and library of files and artifacts.
The approach places Claude Code alongside other tools designed to coordinate groups of AI agents. Rather than handling a single task at a time, the feature is built around parallel workflows, with each project able to progress several strands of work simultaneously.
How the Projects feature works
Within each project, tasks are organised into threads that run different jobs in parallel. A dedicated coordinator directs the activity across those threads, keeping the overall work organised and ensuring that separate tasks remain aligned with the project’s shared aims.
Under the hood, each thread operates as a Claude Code cloud session working on its own branch and its own copy of the repository. This structure means individual agents can make changes independently without immediately affecting the work carried out elsewhere in the project.
Handling overlaps and delegated work
Because threads work on separate branches, situations can arise where more than one thread edits the same code. When that happens, the overlap is resolved as a merge conflict, handled in the same way as any other pull request. This mirrors established version control practices familiar to developers working with shared repositories.
The system also supports further division of tasks. Each thread can split its delegated work into additional strands, allowing a single assignment to be broken down and distributed as required. The coordinator remains responsible for keeping the various strands in order as they progress.
By combining shared memory, parallel threads, and a coordinating layer, the relaunched Projects feature is positioned as a way to manage more complex, multi-step tasks within a single workspace. The reliance on branches and copies of the repository keeps each agent’s work contained, while the merge conflict process provides a structured means of reconciling changes when threads intersect.
Source
Image: theverge.com
