Anthropic Redesigns Claude Code Projects to Coordinate Agent Threads

Anthropic on September 17, 2026, announced a redesigned Projects experience for Claude Code, releasing the update in beta with a coordinator that directs parallel agent threads drawing on shared memory and a common library of files and artifacts.

In the announcement, published on Anthropic’s Claude blog under the title “Projects redesigned: from folder to conversation,” the company said managing multiple sessions across a build used to require users to split the work themselves, manage handoffs manually, and recombine the results at the end. In the updated experience, the user describes what needs to get done and Claude manages the work. Anthropic said Claude scopes the request, delegates tasks, coordinates the parallel threads, reviews outputs, and assembles the finished result. Users can steer progress throughout, including from a phone, and the company said work continues after the user steps away from the computer.

Beta access began September 17 for a select group of Claude Pro and Max subscribers: those who use cloud sessions in Claude Code and have no existing projects on the web or desktop. Anthropic said access will widen over the coming week to more Claude Code users on those plans, with the updated projects reaching the rest of Claude and the Team and Enterprise plans afterward. Pro and Max subscribers who do not yet have access can join a waitlist. Existing projects on Pro and Max plans keep working as they do now, and Anthropic said it will upgrade them as the rollout expands to chat and Cowork.

Goal, Coordinator, and Threads

Under the new structure, projects consist of threads that do the work and a coordinator that directs them. When starting a project, the user selects a goal along with the repository or context, and Claude begins by suggesting work it can pick up right away. Users can configure the project’s cloud environment, connectors, plugins, instructions, and model.

Progress can be monitored and guided in the main project chat, or the user can open an individual thread to examine and steer the details. Anthropic compares the interaction to briefing a chief of staff: the user states the work in the project, and Claude routes it to new or pre-existing threads. Claude also checks in and follows through on assignments. With repositories connected, a thread opens pull requests and runs the project’s tests; with documents, it reads them and produces drafts.

Anthropic’s examples include configuring a project with the goal of reducing an application’s checkout p75 latency, then asking Claude to profile each endpoint, test optimizations, and open pull requests in parallel threads. In another example, the user connects API, web, and mobile repositories and sets a goal of retiring a deprecated v1 endpoint; Claude creates a thread per repository to migrate callers, run tests, and open pull requests, then reports which ones need to merge first.

Cloud Sessions, Branches, and Shared Memory

Each thread is a Claude Code cloud session that works on its own branch and its own copy of the repository. The coordinator keeps the work organized, and when threads touch the same code, the overlap is resolved as a merge conflict, the same as with any other pull request. A thread can also split its delegated work into pieces using subagents, loops, and workflows, which Anthropic said helps large assignments finish faster. The company said projects are built for long-running or agentic workflows, meaning work that runs longer than a single reply and involves more than one part.

Every thread adds to and draws from a shared memory, which Anthropic said lessens the need for complex prompt engineering. Its examples of what Claude can retain include a release moving to Friday, the reason an export was dropped, and the right person to consult before modifying the billing service. Claude also remembers the user’s working and communication style, and users can ask it to adjust how often it checks in, how frequently it starts new threads, and how detailed each update is.

Alongside memory, projects now include a library that collects the files a user adds and the artifacts Claude produces. Anthropic said this makes it easier to locate relevant materials and lets new work build on past efforts.

Usage Limits and What Comes Next

Because each thread is a full Claude Code session and projects can run several threads at once, Anthropic cautioned that projects can reach usage limits faster. Users can check project-specific usage and select the model and effort levels used by the coordinator chat and the worker threads.

Threads currently run in the cloud. Anthropic said the option to run them on the user’s own machine, alongside local tools and code and behind the user’s network, is coming very soon.