Anthropic Adds a Coordinator for Parallel Claude Code Sessions

Anthropic has launched a beta redesign of Claude Code Projects that lets a developer set a larger goal while a coordinator divides it among parallel cloud-based coding sessions. The bet is that developers need help managing the growing number of AI tasks around a build—not just help writing any one piece of code.

The new Projects experience is meant for work with several parts that lasts longer than a single reply. Anthropic says Claude can scope a request, delegate pieces to worker threads, review outputs and assemble the result, while the developer continues to direct the overall project from a main conversation.

One conversation above many coding threads

The setup has two layers. A coordinator receives the developer’s instructions and sends work to new or existing threads. Those threads do the execution: each is a full Claude Code cloud session with its own branch and copy of the repository.

That isolation allows threads to run tests and open pull requests independently. But it does not remove ordinary version-control problems: when agents change the same code, developers still have to resolve the resulting merge conflict just as they would for any other pull request.

Memory is the coordination layer

Parallel workers are only part of the redesign. Anthropic says every thread contributes to and draws from shared project memory, allowing later work to use decisions, requirements and changes in project status without the developer repeatedly rebuilding context. A project library also retains files supplied by the user and artifacts created by Claude.

Anthropic’s examples show the intended scale. A developer trying to reduce checkout latency could ask Claude to profile endpoints, test optimizations and open pull requests in parallel. For a deprecated API endpoint, the system could assign separate migration and testing work across API, web and mobile repositories, then identify the required merge order.

What the developer still controls

• The project goal, connected repository or other context, cloud environment, connectors, plugins, instructions and model settings.
• Whether to monitor work in the main project chat or inspect and steer a specific thread.
• The model and effort level for the coordinator and its worker threads.

Always running, but only in the cloud

Because the worker sessions run remotely, they can continue after a developer closes a laptop. Anthropic says users can follow progress or redirect individual threads from a phone, making the mobile interface a place to supervise work rather than a replacement development environment.

That cloud design is also the product’s immediate boundary. Threads cannot yet run alongside code and tools on a developer’s machine or reach systems behind a private network. Anthropic says local code, tools and network resources are coming “very soon,” without giving a date.

A limited beta with a usage trade-off

The redesigned Projects experience is initially available to select Claude Pro and Max subscribers who use cloud sessions in Claude Code. Anthropic plans to extend access to more people on those plans over the coming week, followed later by updated Projects across Claude and access for Team and Enterprise plans.

There is no separate price for Claude Code Projects. Its coordinator and every worker draw on a subscriber’s existing Claude usage, which means a project running several threads can reach usage limits more quickly. The coordinator can route requests into new or existing threads, but developers still have to resolve merge conflicts when overlapping work changes the same code.