Claude Code Projects is in public beta on Pro and Max, with no separate price and a limit of 200 new threads a day

Jon's agents already run in parallel, and Claude Code is one of the things they run on. Anthropic now offers the coordinating part as a product, Claude Code Projects. Its plan limits, and its reach onto Jon's own machine, decide whether it is worth building on.

It opened in steps. AI Catchup reported that Anthropic opened Projects in beta to some Pro and Max subscribers on September 17. On October 10 a KuCoin news flash reported that everyone on the Pro and Max waitlist had access; that is one outlet's report. As of October 17, Anthropic's documentation says Projects are "in public beta on Pro and Max plans" and "aren't available on Team or Enterprise plans yet."

The same documentation says how it works. A project is one ongoing conversation. Claude starts a thread for each task, usually a cloud session on its own branch, and opens a pull request when the work calls for one. A thread can also run on the user's own computer through Remote Control. The enforced limit is 200 new threads a day across that user's projects. There is no separate price: a project "draws on the same plan limits as your other Claude Code sessions and uses them faster."

This matters because always-on work has needed a machine that stays awake and something to coordinate the sessions. Projects moves the coordinating into Anthropic's cloud and keeps going after the laptop closes. The cost is that cloud threads see only what they are given, and the work comes out of the same plan limits as everything else that runs on Claude Code.

How to use it

1. Check access. Look for Projects in the sidebar at claude.ai/code or in the Code tab of the desktop app. If it is missing, the rollout has not reached the account, and the documentation links a waitlist.
2. Check the code host. Cloud threads need the repository on github.com with the Claude GitHub App installed. GitHub Enterprise Server, GitLab and Bitbucket are out.
3. Create one. New project needs only a name. A goal, repositories and files are optional.
4. Reach a local machine. Run claude remote-control in the folder and leave it running, on Claude Code v2.1.280 or later. In a git repository, add --spawn worktree so each thread gets its own worktree. That thread uses the machine's files, tools, MCP servers, hooks and permission rules. It runs only while the computer is awake.
5. Watch the meter. Project settings has a Usage page with tokens by thread and by model.

The catches in the fine print:

• Cloud threads pick up nothing from the Claude Code setup on the local machine. They get the repository's CLAUDE.md and skills, plus the connectors on the claude.ai account.
• A requested thread count "is a preference rather than a cap."
• A thread that hits the plan limit waits and resumes when the limit resets, so it starts spending the next window without a new message.
• An idle thread watching a pull request wakes up and spends again when CI fails or a review comment arrives.
• If a cloud sandbox cannot be resumed, uncommitted changes can be lost. Have threads commit and push as they go.
• It is not in the terminal CLI, and a project belongs to one user.

What nobody knows yet

• When Team and Enterprise plans get it.
• How much faster it spends a plan. Anthropic gives no figure, only a warning that a Pro plan will reach its limit sooner.
• Whether it stays inside the subscription after the beta.

What to watch

• The plan note at the top of the documentation page. It is where Team and Enterprise access and the end of the beta will show first.