Short answer: Claude Code Projects is one coordinator chat that starts parallel cloud threads, each on its own branch, with shared project memory. It is a public beta on Pro and Max only, and it spends your normal plan limits.
The feature sits inside Claude at claude.ai/code, the desktop app's Code tab and the Claude mobile app for iOS and Android. It is not in the terminal CLI, the VS Code extension or the JetBrains plugin. The useful detail: the coordinator runs at low effort, while every thread runs Opus at high effort.
How the coordinator and threads work
Anthropic's launch post puts it in one line: "Projects have threads that do the work and a coordinator that directs them."
- You brief the project conversation. That chat is the coordinator.
- It starts threads. Each one "is a Claude Code cloud session working on its own branch and copy of the repo."
- Threads open pull requests, which collect in the Pull requests tab.
- The Threads tab groups work by state: Ready for review, Waiting on you, Working, Landing, Idle and Resolved.
The other Overview tabs are Library (files you added and files threads produced), Pull requests and Scheduled tasks. Scheduled tasks was called Routines at launch.
What shared memory holds
Every cloud thread reads an index file, MEMORY.md, when it starts. It opens the other memory files only when it needs them.
Project instructions are a separate layer. They are text "sent to each new thread and to Claude in the project conversation, up to 16,000 characters."
Neither layer is your local setup. The docs say project memory is separate from the auto memory Claude Code keeps on your machine, even though both use a MEMORY.md index. It is also separate from the CLAUDE.md files in your repos.
See it yourself: you can run one thread on your computer with Remote Control
Cloud threads work on GitHub repos and on files, folders and Google Drive folders you upload. When a task needs something only your machine has, you can ask Claude to run that thread on your computer through Remote Control.
- In the desktop app, open Settings > Claude Code, turn on "Use this computer from your phone and claude.ai" and add the folder. Or run
claude remote-controlin that folder and leave it running. - In the project conversation, choose Work locally from the + menu beside the message box.
Three catches. The thread pauses while the computer sleeps and stops if the app or claude remote-control quits. It fails while "Require trusted devices" is on for your account. And you can't add a session you started yourself on your machine to a project.
What it costs and who gets it
The docs we read list no separate price. Threads and the conversation draw on your plan's five-hour and weekly limits. The docs warn that the Opus-at-high-effort default "draws on your plan fastest." For plan prices, check the vendor page.
| Plan | Claude Code Projects today |
|---|---|
| Free | Not in the beta (it covers Pro and Max) |
| Pro | Public beta, rolling out gradually |
| Max | Public beta, rolling out gradually |
| Team | Not yet; "will follow" |
| Enterprise | Not yet; "will follow" |
If Projects isn't in your sidebar, the rollout hasn't reached you. Join the waitlist. The rollout starts with accounts that have used cloud sessions and have no existing projects in claude.ai chat or Cowork.
At a limit, a thread or the conversation retries and continues after the reset. A thread started by a scheduled task doesn't wait; its turn stops with a limit error. A thread cap you type in chat isn't a hard cap either. The docs call these "instructions Claude keeps to, not enforced settings."
Limits to know before you rely on it
- GitHub.com only: not GitHub Enterprise Server, GitLab or Bitbucket. You need push access and the Claude GitHub App installed.
- Org installs: only an organization owner can finish the install. SSO orgs must reconnect GitHub and authorize the Claude app.
- Single user: you can't share a project or its threads. There are no organization-level controls during the beta.
- Uncommitted work can vanish: if a paused sandbox can't resume, the thread restarts from a fresh clone.
- Upload cap: up to 10 folders and Google Drive folders combined per project.
Claude Code Projects vs classic Projects, Codex and Jules
The name collides with the classic Projects in Claude chat. Those are open to free users, who can create up to five. On Team and Enterprise they can be shared with your organization.
| Product | Where the work runs | Vendor's own wording |
|---|---|---|
| Claude Code Projects | Cloud threads, or one thread on your computer | "threads that do the work and a coordinator that directs them" |
| Classic Claude Projects | Claude chat | Free users "can create a maximum of five projects" |
| OpenAI Codex | Cloud environments per repo | "runs many agents in parallel" and "keeps going when you close your laptop" |
| Jules | A Cloud VM | "fully async, multi-agent development"; higher tiers list "60 concurrent tasks" |
For wider context, see our agentic coding tools compared and the full list of AI coding assistants.
FAQ
Is Claude Code Projects available on Team?
No. The Claude Code docs on Projects say it isn't on Team or Enterprise yet. Anthropic says those plans come later.
Can a thread use my local files?
Yes, through Remote Control: Claude can run a thread on your computer instead of the cloud. It runs only while that computer is awake.
Does it cost extra?
The docs we read list no separate fee. It spends your plan limits, fastest on the Opus high-effort default.
Is project memory the same as CLAUDE.md?
No. Project memory is separate from repo CLAUDE.md files and from local auto memory.
Do cloud threads keep running when my laptop is closed?
Yes. Cloud threads keep going after you close your laptop.
Can I hard-cap the number of threads?
Not with a chat instruction. Claude follows it, but it isn't enforced.
When to use cloud vs a local thread
Default to cloud threads: they survive a closed laptop and run in parallel. Move a thread local only when it needs a tool or file that lives on your machine. If your code isn't on github.com, or a team must share the project, wait or look at other agent orchestration tools. For audited pipelines, our guide to governed multi-agent workflows covers the controls this beta lacks.
