A Folder Becomes a Team: Anthropic Rebuilds Claude Projects Around a Coordinator and Parallel Threads

Photo: Ana Las Heras / Wikimedia Commons (CC BY-SA 4.0)

The Library Media — Science, Technology & Health · AI News & Intelligence Desk

A Folder Becomes a Team: Anthropic Rebuilds Claude Projects Around a Coordinator and Parallel Threads

By Catalina Clarke

Share

Published November 15, 2026 · 5:26 AM PT

Photo: Ana Las Heras / Wikimedia Commons (CC BY-SA 4.0)

A Claude project used to be a folder: a set of files, a set of instructions, and a chat that started from that context each time. In a product post dated September 17, 2026 — ‘Projects redesigned: from folder to conversation’ — Anthropic replaced the folder with something closer to an organisation chart: a coordinator that takes the assignment, a set of parallel threads that do the work, and a shared memory they all draw on. Two months into its beta, the design is worth a close look, because it is the clearest statement any major lab has made about how it thinks agent work should be organised.

How it works

The user describes what needs to get done, and, in Anthropic’s words, Claude manages the work: it scopes the request, delegates it, coordinates parallel threads, reviews the outputs and assembles the finished result. The user can steer progress throughout, including from a phone, and the work keeps going after they step away from their computer.

Each thread is a full Claude Code cloud session, working on its own branch and its own copy of the repository. A thread can split its own assignment further, using subagents, loops and workflows, and with repositories connected it opens pull requests and runs tests. With documents instead of code, the same structure reads and drafts. Every thread adds to and draws from a shared memory, which Anthropic says reduces the need for complex prompt engineering. Users can tell the coordinator how often to check in, how frequently to start new threads, and how detailed each update should be.

Anthropic’s own examples are deliberately ordinary engineering work. In one, a project is given the goal of reducing an app’s checkout p75 latency: Claude profiles each endpoint, tests optimisations and opens pull requests in parallel threads. In the other, a team connects its API, web and mobile repositories and sets out to retire a deprecated v1 endpoint; Claude creates a thread per repository to migrate the callers, run the tests and open the pull requests, then reports which ones need to merge first.

The limits Anthropic states itself

Two honest limits are built into the announcement. The first is coordination: the coordinator keeps work organised, Anthropic says, but if threads work on the same code, the overlap is resolved as a merge conflict just like any other pull request. Parallel agents do not abolish the oldest problem in parallel work; they surface it in the usual place.

The second is cost. A project can run several threads at once, and each one is a full Claude Code session — so, Anthropic warns, projects can reach usage limits faster. Users can check project-specific usage and choose the model and effort level for the coordinator chat and for the worker threads separately. Five threads are five sessions drawing on one plan, and the arithmetic of a team applies: the output multiplies, and so does the bill.

Who has it

At launch, Anthropic made the updated projects available in beta to select Claude Pro and Max subscribers who use cloud sessions in Claude Code and do not already have projects on the web or desktop, with expansion to more Claude Code users on those plans over the following week, and updated projects across the rest of Claude, plus Team and Enterprise plans, after that. Existing projects on Pro and Max plans keep working as they do today and are upgraded as the rollout reaches them. Threads run in the cloud today; running them on the user’s own machine, alongside local tools and code and behind their network, is, in Anthropic’s words, coming very soon.

What the redesign signals is a bet about shape. The older idea of an AI assistant was a single conversation that got longer. The project, as Anthropic now builds it, is a small organisation: one coordinator with the whole picture, workers with bounded assignments, a shared memory instead of a longer prompt, and a human who sets goals and reviews results rather than typing every step. Whether that organisation outperforms a room full of everyone’s personal agents is a live question — and one the research is beginning to test directly.

Catalina Clarke
AI News & Intelligence Desk, The Library Media

From The Library — the estate of our family.

Follow Us

X Facebook Instagram Threads Mastodon Bluesky

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top