Claude Code's new Projects bring multi-agent orchestration, but only for cloud users
Anthropic's redesigned Claude Code Projects coordinate parallel threads, share memory, and span multiple repositories, aiming to tame complex AI coding workflows. The feature is initially available only to select Pro and Max subscribers using cloud sessions, with local terminal-based development excluded for now. Anthropic says broader rollout and upgrades for existing project users will come later.
The redesign introduces a coordinator layer that scopes requests, delegates tasks across parallel threads, and assembles finished results, with each thread operating as a full cloud session containing its own branch and repository copy. Named threads allow mid-flight steering, so developers can redirect work without waiting for completion. Anthropic frames the briefing process as similar to instructing a chief of staff, with Claude routing tasks to new or existing threads while accumulating project knowledge over time.
The rollout is deliberately narrow: only select Pro and Max subscribers using cloud sessions, with no existing web or desktop projects, qualify initially. Existing project users keep their current setup until a later upgrade, and local terminal-based workflows remain entirely outside the new system for now.
This feature could meaningfully shift how professional developers coordinate AI-assisted work, particularly for teams managing large, multi-repository codebases where agent sprawl has become a practical problem. Cloud-only availability may create a two-tier experience, potentially pushing more developers toward cloud environments while leaving local-first practitioners with fewer orchestration tools. As Anthropic expands access, the broader impact will depend on reliability and cost—if coordination reduces wasted tokens and failed runs, it could improve productivity; if not, it may simply add another subscription layer to an already complex AI tooling ecosystem.