Claude Code Projects: multiple AI workers on one task

What's new

Anthropic has launched Projects in Claude Code, its tool for AI-assisted programming. The beta began on September 17, 2026, for a select group of Pro and Max subscribers. Pro and Max users who don't have access yet can now join a waitlist while Anthropic gradually opens up to more people. Anthropic hasn't said exactly when everyone on the list will be in.

The first users selected run Claude Code in the cloud (cloud sessions) and don't already have Projects in Claude on web or desktop. Once the feature is turned on, it appears in the sidebar at claude.ai/code or under the Code tab in the desktop app. Team and Enterprise customers, as well as regular Claude chat and Cowork, will follow later. There is no date for that.

What's clever about it

Think of a project as a project manager with a small team. You hand over a big task, and Claude breaks it down, delegates the pieces to multiple threads, reviews the results and pulls everything together. Each thread is a full Claude Code session with its own copy of the code and its own branch, so it can edit code, run tests and open pull requests (proposed changes that others can review).

The project has a shared memory, so decisions and preferences carry across threads. You can steer the work in the project's chat or look in on a single thread. The threads can also split the work up further on their own. You can also choose the model and effort level for both the coordinator and the workers, and track usage for each individual project.

Cheaper or better?

There is no separate price. Projects draws on the same usage allowance you already have in Pro or Max. That means you don't pay extra, but it also means you may hit the limit sooner. Anthropic itself warns that projects can reach the usage limit faster because multiple threads run at the same time, and each one is a full Claude Code session. Anthropic hasn't published how much faster.

Existing chat projects work as before and still count toward your usage. The sources don't make a direct comparison with competitors' similar tools. The monthly price for Pro and Max isn't given in the sources either.

There are also limitations. At launch, the feature ran only in the cloud, not locally on your own computer. If two threads change the same parts of the code, ordinary merge conflicts can arise, and a person has to resolve them.

What it's good for

The feature suits larger tasks that can be split into parts that don't get in each other's way. Examples include updating several modules, writing tests for many parts of an app or cleaning up a large codebase. Here, multiple threads can work side by side instead of one after another.

For small fixes, it's overkill. The sources recommend starting with one well-defined project with clearly separated tasks and only then scaling up. If you haven't been selected yet, you can join the waitlist and use your regular chat projects in the meantime.