Anthropic Launches Claude Code Projects in Beta: Parallel Cloud Sessions That Keep Running After You Close Your Laptop
MarkTechPost Michal Sutter ● Covered by 3 sources
Anthropic turned Claude Code Projects into a coordinator that can spin up parallel cloud threads. They keep working after you close your laptop, and each one can open a pull request.
Based on reporting by MarkTechPost, Michal Sutter — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has redesigned Claude Code Projects, and the old idea of a project as a folder with a chat attached is gone. The new version is closer to a single running conversation that can fan out into work on its own. You describe the task, Claude decides what should become a thread, and those threads keep going in the cloud on separate branches with separate copies of the repository.
Anthropic describes the setup as one conversation that splits itself into parallel cloud sessions. The project conversation is the coordinator. It handles quick questions in place, saves memory, and launches threads when the job needs real work. It does not watch every step. It only sees what the workers report back.
The threads do the heavy lifting. A thread can open a pull request when needed, then watch that pull request with auto-fix turned on. If CI fails, it pushes fixes. If checks pass, it replies. And threads can delegate too, splitting their own assignments further with subagents, loops, and workflows.
Anthropic’s examples are the kind of thing this setup is meant for. One asks Claude to reduce checkout p75 latency by profiling endpoints, testing optimizations, and opening pull requests in parallel. Another retires a deprecated v1 endpoint across API, web, and mobile repositories, one thread per repo, while Claude reports which pull requests merge first.
There is still a very old-fashioned failure mode hiding underneath all this automation: if two threads touch the same code, the overlap shows up as a plain git merge conflict. The future, as ever, eventually runs into git.
My take — AI-written commentary, not fact-checked reporting
This is the right direction, because the real bottleneck in coding with AI is coordination, not just token generation. Anthropic is basically admitting that a single chat window is too small for serious work, which is refreshingly unmagical. The open question is whether teams want an AI coworker that can fan out fast, or just one that leaves fewer messes in git.
Read more about this at: MarkTechPost