Anthropic launches Claude Code Projects for multi-agent AI development

Anthropic has launched the beta version of Claude Code Projects, a new system designed to function as an intelligent orchestrator for multi-agent AI software development. The system utilizes a central coordinator agent that analyzes developer objectives and decomposes them into multiple parallel workstreams.

Each workstream operates as an independent cloud session with its own dedicated Git branch and repository copy. This architecture allows for the simultaneous execution of tasks such as coding, testing, and documentation. To maintain context across these parallel processes, the system employs a shared memory feature—utilizing a MEMORY.md file and instructions up to 16,000 characters—and a central library to collect project artifacts.

While the AI handles task distribution, human developers retain control over the process. The system is designed to trigger merge conflicts rather than silently overwriting code, requiring manual arbitration by the developer when overlapping changes occur. The beta is initially available to select Pro and Max subscribers using cloud sessions, with plans for future expansion to Team, Enterprise, and local modes.