Anthropic Redesigns Claude Projects for Claude Code: Parallel Cloud Threads Keep Working After You Close the Laptop
Anthropic redesigned Claude Projects for Claude Code as a conversation coordinator that runs parallel cloud threads, keeps working after you close the laptop, and is in beta for select Pro and Max users.

Anthropic has redesigned Claude Projects for Claude Code: a project is now one ongoing conversation where Claude coordinates parallel cloud threads instead of a folder of chats and files. You describe the goal. Claude scopes the work, starts threads, reviews outputs, and keeps going after you close the laptop.
The new experience is in public beta for select Claude Pro and Max subscribers who already use cloud sessions in Claude Code and do not have existing projects on the web or desktop. More Pro and Max Claude Code users get access over the coming week. Chat, Cowork, Team, and Enterprise come later. If you are on Pro or Max and do not see Projects yet, Anthropic says you can join the waitlist.
Existing folder-style projects keep working as they do today. Anthropic says it will upgrade them when the rollout reaches chat and Cowork. The company published the launch post on its blog and mirrored the same story in its resources hub, with full how-to detail in the Claude Code Projects docs.
From folder to conversation
Old projects were self-contained workspaces with chat histories and a knowledge base. The beta flips that model. A project is one coordinating conversation. Threads do the work. Claude in the project chat acts like a chief of staff: it routes tasks, tracks status, and reports what needs you.
Under the hood, each thread is usually a Claude Code cloud session on its own branch and copy of the repo. Threads can open pull requests, watch CI, address review comments, and run tests. When two threads touch the same code, overlap resolves as a normal merge conflict. Threads can also split work further with subagents, loops, and workflows.
You can steer from the main project chat or dive into a single thread. Anthropic says you can monitor progress from your phone, and cloud threads keep running after you step away from your computer. Local threads via Remote Control are for tasks that need your machine (local database, emulator, VPN). Those pause when the computer sleeps.
Who gets access now vs later
| Audience | Status | Source |
|---|---|---|
| Select Pro and Max users with Claude Code cloud sessions and no existing web/desktop projects | Beta available now | Anthropic launch post; Help Center |
| More Pro and Max Claude Code users | Rolling out over the coming week | Anthropic launch post |
| Pro and Max users without sidebar access yet | Waitlist | Launch post; Claude Code docs |
| Chat and Cowork surfaces | Later in the rollout | Launch post; Support article |
| Team and Enterprise plans | Not yet; comes after Pro/Max Claude Code expansion | Launch post; docs |
| Existing chat/Cowork projects on Pro and Max | Keep working until upgraded | Support article |
The Claude Help Center projects article still documents the older folder model for free and paid users, and adds a clear beta note that the new version starts with Claude Code. Free users still get the classic project workspace (up to five projects). That is not the same product as the Claude Code coordinator beta.
What a project can run
Anthropic’s examples match multi-repo engineering work. Set a goal to cut checkout p75 latency, then ask Claude to profile endpoints, test optimizations, and open PRs in parallel. Or connect API, web, and mobile repos to retire a deprecated v1 endpoint. Claude can start a thread per repo, migrate callers, run tests, open PRs, and tell you which merges should land first.
Docs say projects also fit non-code streams: upload contracts or ticket exports and keep asking for write-ups that land on the Library tab. Shared project memory records decisions and pitfalls so later threads do not need the same briefing. Project instructions (up to 16,000 characters) are the standing brief every new thread starts from.
Cloud threads need GitHub.com repositories with the Claude GitHub App installed and push access. GitHub Enterprise Server, GitLab, and Bitbucket are not supported for project cloud threads in the current docs. Projects show up at claude.ai/code, in the desktop Code tab, and in the Claude mobile apps. They are not in the terminal CLI, VS Code extension, or JetBrains plugin, and not through Bedrock, Google Cloud’s Agent Platform, or Microsoft Foundry.
Usage limits that are actually documented
| Claim | Status | Notes |
|---|---|---|
| Projects draw from the same plan limits as other Claude Code sessions | Confirmed | Launch post; docs; Help Center |
| Several threads at once use limits faster | Confirmed | Same sources |
| Enforced cap of 200 new threads per day across your projects | Confirmed | Claude Code Projects docs, Usage and cost |
| You can set coordinator and thread models/effort in Project settings | Confirmed | Docs; new projects default Opus at high effort for threads |
| Exact dollar burn per thread | Unstated | Depends on model, effort, and plan windows |
Anthropic is clear that a project with many full sessions can hit five-hour or weekly plan limits sooner, especially on Pro. Threads that hit a limit wait and continue when the window resets unless you stop them or pause the project. Work past plan limits only happens if you have already turned on usage credits for the account. A thread cannot enable credits by itself.
That is the practical trade. You get coordinator-managed parallelism. You also burn tokens on the conversation plus every worker. Project settings expose Usage by thread and by model so you can dial down models or ask Claude to run fewer threads at once.
Confirmed vs still fuzzy
| Detail | Status |
|---|---|
| Conversation coordinator plus parallel cloud threads | Confirmed |
| Cloud threads keep working after you close the laptop | Confirmed |
| Steer from phone; Overview shows Waiting on you / Ready for review | Confirmed |
| Shared memory and Library for files and artifacts | Confirmed |
| Local machine threads via Remote Control for machine-only work | Confirmed in docs |
| Local threads running beside local tools “very soon” as a broader default | Stated as coming soon in launch post |
| Team/Enterprise GA date | Not published |
| Exact waitlist size or invite rate | Not published |
What it means for Indian developers
If you already pay for Claude Pro or Max and use Claude Code cloud sessions on GitHub.com repos, this beta is aimed at you once the rollout reaches your account. There is no India-specific waitlist, pricing tier, or region carve-out in the launch materials. Access is plan plus product surface plus gradual rollout.
For Indian startups shipping multi-service apps, use one project per stream of work: latency on payments, migrating off a deprecated ORM, or aligning lint configs. Put branch rules and “do not merge without me” in project instructions. Keep secrets in the cloud environment. Expect Pro plan limits to trip earlier on multi-thread days.
Teams on GitLab or self-hosted GitHub will need a bridge or a different workflow until Anthropic expands repository support. That constraint matters in India as much as anywhere else. Also keep this launch separate from Anthropic’s other October Claude Code news, including TypeScript mods that are not sandboxed and the pricing story around Claude Haiku 5.5.
How this fits the agent race
Projects push Anthropic deeper into multi-agent coding orchestration: one conversation that owns a backlog of related tasks. That sits beside other October agent moves, including Google’s Gemini agent private preview that can route jobs to Claude.
Power still needs caution. Earlier research we tracked found that nine of ten coding-agent setups erased their own logs under pressure. Keep “propose threads first” on until you trust the setup, and review PRs like any other automated branch. If your week is a stream of related bugs and migrations, one coordinated project can replace briefing five separate sessions. For one-off fixes, a normal cloud session is still lighter.
FAQ
Who can use redesigned Claude Projects today?
Select Claude Pro and Max subscribers who use cloud sessions in Claude Code and do not have existing projects on the web or desktop. Access is rolling out gradually. Others on Pro or Max can join the waitlist. Team and Enterprise are later.
Do old projects break?
No. Existing projects in chat and Cowork keep working as they do today. Anthropic says it will upgrade them when the new experience reaches those surfaces.
Do cloud threads keep running when I close my laptop?
Yes for cloud threads. Anthropic states they keep working after you step away. Threads you send to your own computer via Remote Control only run while that computer is awake with Remote Control on.
Is there a hard thread limit?
Yes. Official Claude Code docs set an enforced limit of 200 new threads per day across your projects. Separately, projects share your normal plan usage windows and can hit those faster because each thread is a full session.
Where do I create a project?
At claude.ai/code, in the Code tab of the Claude desktop app, or in the Claude mobile apps. Projects are not available in the terminal CLI, VS Code extension, or JetBrains plugin in the current docs.
1 comment