Anthropic Turns Claude Code Projects Into Parallel Cloud Session Orchestrator

On September 17, 2026, Anthropic announced the beta launch of a revamped Claude Code Projects platform. The update redefines a “project” as a persistent conversation that automatically creates separate threads whenever a task requires its own execution environment. Each thread runs as an individual Claude Code session in the cloud, complete with its own Git branch and a private copy of the code repository.

The design goal is to eliminate the manual juggling of multiple terminals, local branches, and concurrent AI agents. By delegating each thread to a cloud instance, developers can start a long‑running build, test suite, or data‑processing job and then close their laptop without interrupting the work. The coordinator—essentially the main conversation window—receives concise summaries from each thread and allows the user to answer brief follow‑up questions directly, without opening a new session.

How Parallel Threads Operate

When a developer issues a command that requires isolation—such as compiling a new feature branch or running a resource‑intensive model—the platform spawns a new thread. That thread receives its own branch, ensuring that file changes do not collide with work happening in other threads. The thread continues to execute in the Anthropic cloud even if the user’s local machine is powered down, and it reports progress back to the coordinator as soon as new information is available.

Because each thread maintains a separate copy of the repository, the risk of immediate file‑level conflicts is reduced. However, Anthropic’s documentation makes clear that merge conflicts can still arise when branches are eventually reconciled, and developers must still review changes before integration.

Current Availability and Limitations

The parallel‑session feature is, for now, limited to the web‑based and cloud version of Claude Code. Users who rely exclusively on a local installation must either keep their existing workflow or wait for future updates that bring cloud orchestration to the desktop client. This staged rollout allows Anthropic to gather performance data and user feedback before expanding the feature set.

Anthropic does not claim independent, quantified productivity gains. Instead, the company frames the new orchestration as a way to reduce the cognitive load associated with managing multiple terminals, branches, and agents, promising a smoother development experience.

Key Functional Changes

• Automatic creation of cloud‑hosted threads for each distinct task
• Dedicated Git branch per thread to isolate file changes
• Continuous execution after the user’s device is closed
• Coordinator view that aggregates thread summaries and accepts brief replies
• Web‑only availability during the beta phase

The coordinator’s summary feed is designed to surface only the most relevant information—such as build success, test failures, or model output—so developers can stay informed without switching contexts. If a thread encounters an error that requires a decision, the coordinator prompts the user with a short question, allowing a quick response without opening a separate Claude Code window.

From an operational standpoint, the architecture leverages Anthropic’s existing cloud infrastructure. Each thread runs in an isolated container, preserving security boundaries while sharing the same project metadata. This approach also enables Anthropic to scale resources dynamically, allocating more compute to threads that need it and releasing it when they finish.

Early adopters have reported that the ability to close their laptops and let long‑running jobs finish in the background eliminates a common source of interruption. Teams can now start a heavy data‑processing job in the morning, walk away for meetings, and receive a concise status update in the coordinator by afternoon.

What this means for an English‑speaking organization is a tangible reduction in the time spent switching between local terminals, managing branch merges manually, and monitoring idle machines. By centralising orchestration in a single web interface, teams can keep development momentum while freeing up laptops for other tasks, ultimately allowing engineers to focus more on design decisions and less on tooling overhead.