What happened

On November 14, Anthropic shipped an update that turns Claude Code from one long chat that writes code into a small team of parallel cloud sessions. You no longer feed a project in file by file and babysit a single thread. You describe the job once — say, 'add user accounts, write the tests, refresh the docs' — and a coordinator takes it from there. It scopes the work, splits it across separate threads, reviews what comes back, and assembles the finished result.

Each thread is a full Claude Code cloud session with its own copy of your repository and its own Git branch. Two quick translations for the rest of us: a repository is just your project folder plus a complete history of every change ever made to it, and a branch is a private scratch copy where edits can't break the original. A thread can go deeper still and break its slice into subagents, loops or workflows — think of a team lead handing parts of a task to junior helpers.

The work keeps moving after you shut the laptop

That's the real pitch. You can close the lid, walk away and check in from your phone while the actual labor fans out in the background. Steering stays in one conversation; the doing happens elsewhere.

Before you get too excited, though: this is a public beta. It's rolling out gradually to selected subscribers on the Pro, Max and Ultra plans who already use Claude Code cloud sessions. Team and Enterprise plans don't have it yet, and there's a waitlist for everyone else. So availability right now is a lottery, not a product page.

What it means for you

The shift is easier to picture with a non-tech analogy. Old Claude Projects were like a filing cabinet you could talk to: you opened a drawer, pulled out a file, asked a question. The new version is like hiring a project manager who reads the cabinet, writes the to-do list themselves, farms out the pieces and hands you one finished folder.

You still don't need to write code yourself. You need to be able to describe an outcome in plain English — and that's a skill most people already have.

At home

Picture a family budget tracker that pulls numbers from your bank exports, flags weird spending and emails you a weekly summary. Normally that's a weekend of Googling. Here you describe the whole thing in one go, and separate threads can handle the import, the categorizing and the email part at the same time.

At work

Let's say you're a freelancer with three clients and a messy shared folder. One thread tightens up a project's documentation, another cleans the data files, a third writes the tests that catch mistakes. You review on the train home instead of at your desk at 11pm.

For a small business

A shop owner who needs a product page, a stock-check script and a short 'how to order' guide can describe all three at once. The coordinator keeps them on separate branches, so a broken script can't take down the page.

For studying

A student building a data-analysis project for a thesis can push the cleaning, the charts and the write-up of the method into different threads. Each one works on its own copy, which is a lot less scary than editing the only version you have.

For creative and income projects

This is where it gets interesting for people who don't think of themselves as technical. A newsletter, a small app, a paid template pack — all of it is code plus text, and all of it can be split up. The bottleneck stops being 'I can't build this' and becomes 'I have to review this.'

Whichever scenario sounds like yours, the raw material is a clear written request. If you want help sharpening that request, free AI tools such as the ones at MyKreaTool can tighten a messy brief in a couple of minutes.

How to try it right now

Step 1: Check whether you're even in the room

Log into your Claude account and open Projects, then confirm two things: your plan is Pro, Max or Ultra, and you already use Claude Code cloud sessions. If you're on Team or Enterprise, it's not there yet. If you're on anything else, get on the waitlist and keep an eye on your inbox.

Step 2: Start free — write the brief before you spend anything

Don't burn usage limits learning how to ask. Draft your job description in a free tool first, cut it down to one paragraph, and make sure it names the outcome, not the technique. 'A page where customers can filter products by size' beats 'build a React component with useState.'

Step 3: Open a Project and describe the whole job once

Put your files in, then write the single conversation message you'd give a new hire on day one. Mention what 'done' looks like. The coordinator does the splitting, so you don't need to plan the threads yourself.

Step 4: Let it run, then steer from your phone

Close the laptop. Come back later and read the coordinator's summary rather than every thread. If one piece went sideways, redirect in the main conversation instead of diving into the branch.

Step 5: Review pull requests before merging

A pull request (PR) is simply a proposal to add someone's changes to the real project. Read each one, then merge. Anthropic's built-in Projects documentation walks through the mechanics if you get stuck.

Step 6: Keep the threads small at first

Start with two or three parallel threads on genuinely separate parts of the job. Fewer threads means fewer collisions and a much easier review.

Upsides and what changes

The obvious win is throughput on work that was never really one job. Importing data, writing docs and fixing tests don't need to happen in sequence — they were only sequential because one chat had to do them one at a time. Spreading them across parallel cloud sessions with separate branches means the slow parts overlap instead of queueing.

The second win is the phone. Handing off a task at 9am and checking the summary at lunch is a different way of working, and it's the part most people will actually feel day to day.

The honest open question — and it's the one the original poster asked — is whether parallel agents speed up real work or just move the bottleneck downstream to review. Splitting one task into four doesn't shrink the reading you have to do at the end. If anything, you now have four outputs to check instead of one. The bet Anthropic is making is that generating work is the expensive part; for a lot of small projects, checking it might turn out to be.

Limitations

Anthropic is refreshingly blunt about the rough edges. Threads run only in the cloud for now — if your code lives on your own machine behind a VPN, you'll have to wait, though local code and VPN support is described as coming 'very soon.' Several parallel sessions burn through your usage limits faster, which is worth remembering before you launch seven threads on a Tuesday morning. If two threads touch the same piece of code, the clash shows up as an ordinary merge conflict in a pull request — the familiar 'which version wins?' moment, not a magic fix. And the whole thing is a gradual beta limited to selected Pro, Max and Ultra subscribers, with Team and Enterprise waiting and everyone else on a waitlist.

Conclusion

Parallel cloud sessions are a real change in how Claude Code works: one description in, a coordinated set of branch-isolated threads out, still running while your laptop is shut. The bottleneck has moved to review — and that's fine, as long as you plan for it. Your one action today: write a one-paragraph brief for a project you've been putting off, name the outcome, and see whether it splits cleanly into separate threads. If it does, you're ready for the beta the moment you get access.