On 17 September 2026, Anthropic rebuilt Claude Code Projects around a coordinator model: instead of a single conversation tied to a folder of files, Projects now scopes an incoming request, splits it across multiple parallel threads — each one a cloud session with its own branch and its own copy of the repository — and reviews and assembles what comes back. Threads share memory and a common library of files and artifacts, and work keeps running in the cloud after you close the laptop; progress can be steered from a phone. It’s in beta for select Pro and Max users, with wider rollout to follow.
The shift is from “an assistant” to “a team you don’t have to manage”
Most AI coding tools to date have sold single-threaded speed: one assistant, one task, faster than a human would do it alone. A coordinator that delegates across parallel threads is a different claim — it’s managing concurrency, not just throughput, which is the part of running a dev team that actually consumes a technical lead’s time. Whether that holds up under real project complexity, rather than benchmark tasks, is the open question with any newly launched agentic feature; the pattern across 2026 has been that these capabilities improve fast once real usage starts surfacing edge cases.
So what
For anyone evaluating whether to commission a build with an AI-assisted team versus a traditional one, this is a preview of where the tooling is heading: less “does the AI write good code” and more “can the coordination layer be trusted to run several workstreams without a human catching every handoff.” That’s exactly the gap between speed and reliability we build review and accountability around — see our AI-assisted development approach, or get in touch if you want a project scoped with that distinction in mind from day one.