Anthropic has launched an upgrade to Claude Code Projects, introducing the ability to orchestrate parallel AI workers across complex software development tasks. Under the new architecture, users define a high-level technical objective, while a central Claude coordinator scopes requirements, breaks down assignments, delegates work to separate cloud execution threads, and synthesizes the outputs. Each thread operates in its own isolated cloud environment and repository branch, running tests and opening pull requests independently.
The platform resolves overlapping changes across threads using standard git merge conflict workflows, while allowing individual threads to spin up sub-agents for specialized tasks. In addition, Anthropic introduced shared project memory and a centralized file library. Every worker thread continuously updates and queries this persistent memory, allowing the system to retain architectural decisions, deployment schedules, and user preferences without requiring manual prompt context in each session.
The feature is currently rolling out in beta to select Claude Pro and Max subscribers utilizing cloud sessions, with plans to expand availability to Team and Enterprise accounts. Developers can steer projects continuously or review updates remotely, position modern AI coding from isolated single-prompt interactions toward fully autonomous, multi-threaded codebase maintenance.
Why it matters
Shifts AI coding tools from single-chat assistants to multi-threaded autonomous project managers operating directly on codebases.
Reduces manual prompt context engineering by introducing shared, persistent project memory across developer sub-agents.
Enables faster end-to-end refactoring and bug fixing through parallel cloud execution and automated git branch management.
Source: claude.com



