Summary
Anthropic introduced Claude Code Projects, a beta experience where a single project coordinates multiple parallel Claude Code cloud sessions. Users describe an outcome and Claude scopes the request, delegates work across threads, reviews outputs, and assembles the result, while a shared project memory and a project library of files and artifacts persist across long-running work.
What changed
Claude Code gained a Projects experience that turns folder-based sessions into a conversation-driven workflow: one project scopes a request, delegates it across parallel cloud-session threads (with further subagents and loops when useful), reviews outputs, and assembles the finished result, backed by persistent shared memory and a project library of uploaded files and Claude-generated artifacts.
Why it matters
Coordinating multiple concurrent coding agents has meant manual work-splitting, handoffs, and result-stitching. Building that orchestration and persistent context into the product moves Claude Code from single-session assistance toward managed, long-running project execution, and stakes Anthropic's position in the emerging category of agent 'project' surfaces for sustained software work.
Evidence excerpt
Claude scopes the request, delegates the work, coordinates parallel threads, reviews the outputs, and assembles the finished result.