Anthropic relaunched Claude Code Projects on September 17, 2026, giving users a way to run and coordinate multiple AI coding agents from a single workspace. The feature, now in beta for select Claude Pro and Max subscribers, lets managers assign parallel tasks to different agents while a central coordinator tracks progress and resolves conflicts.
How the multi-agent system works
Each project operates under a shared memory, a common set of goals, and a unified library of files and artifacts. Individual tasks run in separate "threads" - each one a Claude Code cloud session working on its own branch and copy of the repository. A coordinator agent directs the threads and keeps work organized. If two threads modify the same code, the overlap is handled as a standard merge conflict, the same way a pull request would be resolved.
Threads can further break down delegated work using subagents, loops, and workflows. This means a large assignment can be split into smaller pieces that run simultaneously, reducing total completion time. Users can interact with any thread individually or monitor everything through the main project chat. At launch, threads run exclusively in the cloud, but Anthropic said support for local tools and code is coming "very soon."
Availability and rollout
The updated Projects feature entered beta on September 17 for a limited group of Claude Pro and Max subscribers. Anthropic plans to expand access to all Pro, Max, Team, and Enterprise users later, along with Cowork and regular Claude chats. The company did not specify a timeline for the broader rollout.
The approach mirrors tools like Grok Bot, which also manage groups of AI agents working in parallel under centralized direction. For teams already using Claude AI Courses & Certifications to upskill, the Projects feature adds a practical layer of orchestration to daily workflows.
Why this matters for management
Managers overseeing technical teams get a coordination model that looks familiar: assign work, track parallel progress, and resolve merge conflicts when agents step on each other's code. The structure reduces the overhead of manually dividing coding tasks and stitching results together. For managers evaluating AI adoption, the feature also signals a shift from single-agent assistants toward systems that mirror how actual teams operate - with delegation, branching workstreams, and centralized oversight. Understanding these coordination patterns will matter as more vendors build multi-agent tooling into enterprise plans, and resources like AI for Management can help leaders assess where such systems fit into their operations.
Your membership also unlocks:
