Draft poster · the finished cut will appear here
What happened
The Verge reports that Anthropic relaunched Claude Code Projects around coordinated cloud work across multiple agents. The report says a project can share goals, memory, files, and artifacts across threads while a coordinator routes the work.
How it works
According to The Verge, each thread runs as a separate Claude Code cloud session on its own branch and repository copy. Overlapping edits surface as ordinary Git merge conflicts, and threads can subdivide work into smaller parallel jobs with subagents, loops, and workflows. The isolation is the key mechanism: workers can proceed independently while the project retains a shared record of the larger objective.
The report says Projects was in beta for a subset of Claude Pro and Max users at publication. Local-tool and local-code support was planned for later, which keeps the first version focused on a hosted development environment.
The branch boundary also gives a human team a familiar review point. Work can be compared, merged, or rejected with Git’s existing mechanics instead of requiring a new hidden synchronization system.
Why it matters
Multi-agent coding gets more tractable when isolation, coordination, shared context, and Git integration are first-class parts of the harness. Each worker gets a space to make changes, while the merge boundary remains visible instead of being hidden inside a coordinator.
What we don’t know yet
The Verge reports that Projects was in beta for a subset of Claude Pro and Max users and that local-tool and local-code support was planned later. The report does not establish how conflicts are resolved at scale, what limits apply to shared memory and artifacts, or when local support will arrive. The described capabilities should not be treated as general availability.
Source: The Verge, Stevie Bonifield, September 17, 2026 - original report