Anthropic Reimagines Claude Projects With Multi-Agent Coordination

Anthropic has redesigned its Projects feature within Claude Code to support coordinated teams of AI agents working on complex tasks in parallel. The new system allows multiple agents to share memory, align on goals, and access a shared repository of files and artifacts—all orchestrated from a cloud-based interface.

How Multi-Agent Coordination Works#

Each Project contains threads, which function as independent Claude Code sessions operating on separate branches of a codebase. A central coordinator oversees the entire team, ensuring work stays organized and preventing bottlenecks. When threads operate on overlapping code sections, merge conflicts are resolved through standard pull request workflows, mirroring traditional development practices.

Individual threads can subdivide their work further using subagents, loops, and workflows—allowing complex assignments to complete faster without centralizing every decision. This hierarchical delegation mirrors how human teams scale responsibilities across multiple contributors.

Current Capabilities and Roadmap#

Users can monitor and direct work through the main project chat, or dive into individual threads for granular control. At launch, all thread execution happens on Anthropic's cloud infrastructure, but the company has pledged to add local tool and code support "very soon."

The revamped Projects feature rolls out in beta immediately to select Claude Pro and Max subscribers. Full access will expand to all Pro, Max, Team, and Enterprise tiers, as well as users of Cowork and standard Claude chats, over time.

Read more about the feature on Anthropic's official blog.