Claude Code Projects Gains AI Orchestration, But Local Developers Are Left Out

Anthropic has overhauled Claude Code Projects, turning the tool from a single-thread coding assistant into an orchestration layer that can coordinate multiple workstreams, share memory between them, and operate across separate repositories. The update, reported by ZDNet, targets developers wrestling with complex, multi-part coding jobs that a single conversation window cannot hold. But the improvements come with a catch: the orchestration features are available only in the cloud version, leaving local users with the older, narrower experience.

The core change is architectural. Instead of treating each task as an isolated chat, Projects now lets a developer group related threads and have them draw on a common context. That shared memory means a fix applied in one thread can inform a refactor in another, and the system can follow dependencies across repositories without the developer manually pasting code back and forth. For teams managing sprawling codebases, that is a meaningful shift from the copy-paste workflows that have defined AI-assisted coding so far.

Why it matters: AI coding tools have largely been judged on single-file completions or isolated bug fixes. Orchestration pushes them toward project-level work, where the hard part is not writing a function but keeping many moving parts consistent. By letting Claude Code hold state across threads and repos, Anthropic is betting that developers will trust an AI system with more of the planning layer, not just the typing. That is a bigger claim than autocomplete, and it puts the company in closer competition with platforms like GitHub Copilot Workspace and Cursor, which have also been adding multi-step agent features.

The cloud-only limitation is the practical snag. Developers who run Claude Code locally, often for privacy, latency, or cost reasons, do not get the orchestration benefits today. That splits the user base into two tiers: cloud users get the coordination machinery, while local users keep the older behavior. Anthropic has not said when or whether local support will arrive, and the company did not offer a timeline in the ZDNet report. For teams with strict data-handling rules, that gap could keep the most interesting parts of the update out of reach.

The release also raises questions about how shared memory is stored and isolated between projects. If context leaks across threads or repositories, the feature could create as many bugs as it prevents. Anthropic will need to show that the memory layer is predictable and that developers can inspect or reset it. Those details will decide whether orchestration feels like a reliable engineering tool or an opaque convenience.

What to watch next is whether Anthropic extends the orchestration features to local installs and how competitors respond. If cloud-only remains the model, the update may accelerate a split between developers who can use hosted AI agents and those who cannot. The company's next move will signal how serious it is about serving the full Claude Code user base, not just the part that lives on its servers.