Claude Code Projects let one Claude conversation split your work into several cloud sessions that run in parallel, each on its own Git branch, and keep track of all of them for you. Anthropic’s ClaudeDevs account said on 9 October that every Pro and Max subscriber already on the waitlist now has access, and that more will be let in as capacity allows. It is still a beta, it is not available on Team or Enterprise plans, and while it costs nothing extra, it eats into your plan’s usage limits faster than a normal session.
That last point matters most if you are on Pro rather than Max. Anthropic’s pricing page lists Pro at $20 a month, or $17 a month when billed yearly at $200, and Max from $100 a month. A project that runs four threads at once is, in usage terms, four Claude Code sessions.
What Claude Code Projects actually do
A project has two parts. The project conversation is where you talk to Claude, which acts as a coordinator. Threads are the workers: each is a separate Claude Code session with its own context window, and a cloud thread works on its own branch and opens a pull request when the work needs one.
In practice, you paste a bug report, a stack trace or a list of tasks into the conversation, and Claude decides what to do with each. A quick question gets answered in place. New work becomes a new thread, or goes to a thread already working on that part of the code. Once a thread opens a pull request, it watches it, pushes fixes when CI fails and responds to review comments.
Three things carry across every thread, so you do not have to repeat yourself:
- Project instructions: a brief of up to 16,000 characters that every new thread starts with, such as which branch to target and which tests to run before calling work done.
- Project memory: notes Claude keeps about decisions and pitfalls, stored as files you can read and edit in Project settings.
- Repositories and files: the GitHub repos every thread clones, plus documents or Google Drive folders you add to the project’s Library.
An Overview pane groups threads by state, including Working, Waiting on you, Ready for review and Resolved. This is the part that saves time: you can leave for an hour, and cloud threads keep running after you close your laptop, so you come back to a list of what finished and what needs an answer.
Who can use Claude Code Projects right now
You need a Claude Pro or Max plan. Anthropic’s documentation says the beta is rolling out gradually, starting with accounts that have used Claude Code cloud sessions and do not already have projects in claude.ai chat or Cowork. If Projects does not appear in the sidebar at claude.ai/code, the rollout has not reached you yet, and the waitlist form is still open. The Free plan does not get Projects, although it now includes Claude Docs, Slides and Design.
Team and Enterprise plans are not included yet. If your Claude access comes through an employer’s Team or Enterprise account, you will have to wait, even if you use Claude Code every day at work.
The beta comes with a few more catches:
- Where it works: claude.ai/code, the Code tab of the Claude desktop app, and the Claude apps for Android and iOS. It is not in the terminal CLI, the VS Code extension or the JetBrains plugin, and not on Amazon Bedrock, Google Cloud or Microsoft Foundry.
- GitHub only: code must be on github.com, with push access from your connected account and the Claude GitHub App installed. GitLab, Bitbucket and GitHub Enterprise Server repos do not qualify.
- Organisation repos: only an organisation owner can install the GitHub App on an organisation’s repositories, so you may need to ask for approval.
- No sharing: a project belongs to one person. You cannot share it or its threads during the beta.
How to start your first Claude Code project
- Open claude.ai/code or the Code tab in the desktop app, select Projects in the left sidebar, then New project.
- Give it a name. Optionally add a one-line goal, such as keeping an API under a latency target, and the GitHub repositories or files it should work on.
- Click Create project. On your first project, Claude takes a turn on its own straight away unless you send a message first, and that turn counts against your plan.
- Open Project settings from the gear icon and write project instructions under Memory: the branch to start from, how to check work, and what needs your approval first.
- Send one small, real task before a big batch, and read how that thread reports back.
If you already have a cloud session doing the work, its menu offers Continue as project, which creates a project from it and suggests a setup.
How fast Projects use your Claude limits
A new project runs Opus everywhere, with threads at high effort, which Anthropic says draws on your plan fastest. You can change the thread model and effort under Project settings, General. Anthropic’s own documentation warns that on Pro you should expect to hit your limit sooner on days you run a project.
- There is no fixed thread count. Claude starts as many as the work needs, and asking it to run only two at a time is a preference rather than a hard cap. The enforced limit is 200 new threads per day across your projects.
- A thread that hits your five-hour or weekly limit keeps retrying and carries on when the limit resets, so work you leave running will start using your next window without asking.
- A finished thread wakes up again if CI fails or a review comment arrives on its pull request.
- An idle project with no running threads, watched pull requests or new messages uses nothing, and neither does an archived one.
- Work goes past your plan’s limits only if you have turned on usage credits. A thread cannot turn them on for you.
Project settings include a Usage view that breaks down token use by thread and by model. If you like seeing your remaining quota at a glance, tools that track AI quota usage are worth a look before you start leaving threads running overnight.
Running a thread on your own computer
Some tasks need a local database, a device emulator or an API behind your office VPN. For those, a project can run one thread on your machine through Remote Control while the rest stay in the cloud. You need Claude Code v2.1.280 or later. In the desktop app, turn on Use this computer from your phone and claude.ai under Settings, Claude Code, and add the folder. From a terminal, run this in the folder and leave it running:
claude remote-controlThen choose Work locally from the + menu in the project conversation. A local thread only runs while that computer is awake, and it does not load the project’s memory files, though it does get the project instructions.
When a project is the wrong tool
A project earns its keep when one goal keeps producing tasks: a migration bigger than one session, a lint change across many repositories, or a service whose bugs you keep feeding in. For a single fix, a normal cloud session is simpler and cheaper. If every task needs your local setup, a local session fits better. And if you prefer a terminal-first agent you can bend to your own workflow, OpenCode plugins and themes give you a lot of that control without a cloud coordinator.
My take
Projects looks most useful for solo developers who juggle several repos, because it removes the busywork of briefing each session and checking which one finished. On Pro, though, I would switch threads off Opus at high effort before the first batch, and set a rule in the instructions that Claude proposes threads before starting them. Max users have more room to let it run.
FAQ
Can I use Claude Code Projects on my phone?
Yes. The Claude apps for Android and iOS support Projects, so you can check on threads, answer the ones waiting on you and send new work while cloud threads keep running.
Can I use Claude Code Projects from the terminal?
Not directly. Projects work at claude.ai/code, in the desktop app and in the mobile apps. The terminal only comes in when you run a thread on your own computer through Remote Control.
Can I share a Claude Code project with my team?
Not during the beta. A project belongs to one user, and Team and Enterprise plans do not have Projects yet.
