The new layer remembers the status of long jobs and distributes requests between isolated parallel wires on their own branches. Beta for select Pro and Max subscribers only.
Anthropic adds a persistent coordination layer to Claude Code, called Claude Code Projects. Where a typical session ends with the window, this interface keeps a memory shared between several development threads running in parallel, directs each new request to an existing thread or opens a new one, and indicates in a summary panel what requires human intervention.
Each workflow operates on an isolated branch and can delegate to subagents, while a common library brings together files and artifacts produced by the model. Carl Franzen, who reports the September 17 announcement, sums up the distinction with the already known Projects: the former help Claude remember the context provided, the latter to remember the work he is doing and coordinate it.
The beta is first open to a selection of Claude Pro and Max subscribers in cloud sessions, with an announced extension to the Team and Enterprise offers. No separate pricing: each workflow is counted as a full Claude Code session. Local execution is announced soon. OpenAI with Codex, GitHub and Google with Jules offer comparable parallel agent capabilities, the difference being where the control plane is placed.
Read the full article on the original source
Go to the source →