Claude Projects Is Back, and It's a Different Thing Now

TL;DR: Claude now has a mode where one conversation hands tasks to several agents working in parallel and tracks them for you. It’s in public beta on Pro and Max, rolling out slowly, and you can join the waitlist. It shares a name with the old Projects but it’s a different feature, and it uses your limits faster.

I already run a small team of bots for this blog, so a coordinator that hands out work is exactly what I’d want to compare. I pay for Claude Max. I haven’t tried it.

What it is

A project is one ongoing conversation. You tell Claude what needs doing, and it starts a thread for each task. Each thread is usually a cloud session, meaning Claude Code running in the cloud instead of on your machine. The threads run in parallel and keep going after you close your laptop.

So you stop being the dispatcher. You set context once, and every new thread starts with the project’s instructions. Anthropic’s example is a rule like which branch to target, stated once and reaching every thread.

Not the old Projects

If you used Projects in the Claude chat app to group conversations and files, this isn’t that. The old version has no threads and no coordinator, and those projects keep working until the redesign reaches them. Ethan Mollick put it as “Same name but no relation to old Projects.”

Who can get it

It’s in public beta on Pro and Max, rolling out gradually, starting with accounts that have used cloud sessions and don’t have existing projects in claude.ai chat or Cowork. It isn’t on Team or Enterprise yet. If it’s not in your sidebar, the rollout hasn’t reached you, and you can join the waitlist on Anthropic’s Projects form.

It works at claude.ai/code, in the desktop app and in the mobile app. I’m not sure if or how it’ll show up in the CLI or the various IDE extensions.

The cost line

This one matters on a flat plan. A project draws on the same limits as your other Claude Code sessions, and the docs say it “uses them faster.” Several agents at once means several agents’ worth of usage.

What I’d test first

Once it shows up, I’d give it the kind of work my own bots do and watch three things.

• Context drift: does a thread still follow the project’s rules after a long run, or does it wander?
• Shared branches: what happens when two threads touch the same branch at once?
• Limits: how fast does a day of parallel work burn through Max?

I don’t know the answers, which is the point of testing.

Projects doesn’t show under Code in my Claude desktop app yet. I’ve never used cloud sessions, and the rollout starts with accounts that have. I’m on the waitlist and plan to try it as soon as I get access.