Anthropic revives Projects as a multi-agent management platform in the cloud

The refreshed Claude Code feature lets you run an entire team of code agents under one roof — shared memory, shared goals, a single library of files and artifacts. Think Grok Bot or other tools that orchestrate swarms of AI agents. Each project spins up "threads" that execute different tasks in parallel, while a "coordinator" keeps the big picture in view.

Under the hood, every thread is its own cloud session of Claude Code, working on its own branch and its own copy of the repo. The coordinator maintains order; when two threads touch the same code, the clash resolves as a standard merge conflict, just like any PR. A thread can further fan out work via subagents, loops, and workflows so large jobs finish faster.

You can chat with each thread individually or oversee and steer everything from the project's main chat. At launch, threads run cloud-only, but Anthropic promises local tool and code support "very soon" — a claim worth taking with a grain of salt until it ships.

The updated feature enters beta today for a select group of Claude Pro and Max subscribers. Access will widen to all Pro, Max, Team, and Enterprise users, plus Cowork chats and regular Claude chats. Anthropic hasn't published benchmarks comparing the multi-agent approach to single-agent work, so it's hard to gauge how much faster "large tasks finish" in practice.