Claude Code Launches Projects to Run Many Threads at Once

Claude Code launches Projects on desktop and web

A project gathers all your related work in one place. You keep dropping work in, whether that’s a bug report, a feature, or a task list, and it splits the work into separate threads on its own. Each thread is a cloud session running in parallel. They keep working even after you close the laptop, and you can check on them or give new instructions from your phone.

An example from the screenshot

In the example, Claude runs several threads at once inside a single project (the checkout-service repo). One thread is updating a dependency, while another has finished fixing a performance bug and already opened a pull request. Anything that needs your decision, such as a breaking change, Claude brings to you, and the Overview pane puts threads waiting on you in their own “Waiting on you” group, separate from finished work.

Worth knowing

• Only select Pro/Max accounts for now: those that have used cloud sessions before and have never had the older Claude Projects in claude.ai chat or Cowork. If you already have old projects, you’ll get access later so your data can move over cleanly. If you don’t qualify yet, you can join the waitlist.
• Where it works: claude.ai/code, the Code tab in the desktop app, and the iOS/Android mobile apps. It isn’t available in the terminal CLI yet.
• A limit of 200 new threads per day per account. A new project runs every thread on Opus at high effort by default, which uses your plan faster than usual. You can lower the model or effort later.
• Threads run only in the cloud for now. They can’t work with files or tools that exist only on your machine. Anthropic says local workflow support is coming soon.

What kind of work it suits

According to Anthropic’s documentation, a project is worth creating when the work has a goal that outlasts one session and keeps producing new tasks, for example:

• One goal across many repositories, such as bringing every service up to a new lint config. Claude can run one thread per repository, each with its own pull request.
• An area you keep feeding, such as the bugs, stack traces, and review requests for one service. A pitfall you tell Claude to remember after one fix carries over to the next thread.
• A build or migration bigger than one session, such as “build what docs/spec.md describes.” The work splits into threads, and decisions you ask Claude to remember early on reach the later ones.
• Work that isn’t code, such as a folder of contracts or a support-ticket export. Upload the documents instead of adding a repository, and threads deliver their results as files on the Library tab.

Other options fit better for a single task that fits in one session (start a cloud session yourself), work that needs a database or tools only on your machine or behind a VPN (use a local session), a task that just runs on a schedule (use a routine), and work several teammates steer together in Slack (use Claude Tag).

What’s inside a project

• The project conversation, where Claude acts as coordinator: it takes what you send, decides what becomes a thread, and tracks every thread it started. It sees what threads report back, not every step they take.
• Threads, each with its own context window, working on its own new branch. Once a thread opens a pull request, it keeps watching it with auto-fix on, pushing fixes when CI fails and addressing review comments.
• The Overview pane, which groups threads as Ready for review, Waiting on you, Working, Landing, Idle, and Resolved, with tabs for the Library, Pull requests, and Routines. The desktop app also sends a notification when a thread hits an error or needs your input.

Set context once for every thread

• Project memory is a set of notes Claude writes itself as it works, such as requirements, decisions, and pitfalls. You can ask it to remember or forget something in the conversation, and edit the files in settings.
• Project instructions are standing text sent to every new thread, up to 16,000 characters. They suit rules like which branch to start from, how a thread checks its own work before calling it done, and what needs your go-ahead.
• Repositories, files, and environment: every thread clones the project’s repositories and loads the CLAUDE.md, skills, and plugins in them. The connectors on your claude.ai account also work in every thread with no per-project setup.

How to get started

At claude.ai/code or in the Code tab of the desktop app, select Projects in the sidebar, then New project. Only the name is required; the goal, repositories, and files can come later. You can also turn a cloud session that’s already doing the work into a project with Continue as a project.

For code work, the code must be on github.com (GitHub Enterprise Server, GitLab, and Bitbucket aren’t supported), your connected GitHub account needs push access, and the Claude GitHub App must be installed on the repository.

The docs suggest writing project instructions first, then sending one small piece of work to see how a thread reports back before sending a larger batch.

About usage

• A project draws on the same plan limits as your other Claude Code sessions and can’t spend past them on its own. The docs warn that Pro users in particular should expect to reach their limit sooner on days they run a project.
• A thread that hits the limit waits and continues on its own when the limit resets, so work you left running starts using your next usage window without a message from you. To avoid that, click Stop in the thread or pause the whole project.
• Threads watching a pull request wake up and use your plan again when CI fails or a review comment arrives. A project with nothing running and no new messages doesn’t use your plan.
• You can see token use by thread and by model in Project settings > Usage.

Limitations during the beta

• It isn’t available on Team or Enterprise plans yet, or through Amazon Bedrock, Google Cloud, or Microsoft Foundry.
• A project belongs to one user. You can’t share a project or its threads, and there are no organization-level controls yet.
• You can’t move a thread to another project.
• A thread’s sandbox pauses between turns. If it can’t be resumed, the thread continues from a fresh clone, so uncommitted changes can be lost. On long tasks, ask Claude to commit and push work in progress.

Editor’s take

To me, what sets this apart from running sessions yourself is that Claude manages the whole work queue for you. You don’t have to keep track of which thread is doing what. It suits work that outlasts a single session and keeps bringing in new tasks, more than one-off jobs you finish in one go.

---

The information in this article is based on the official Claude Code documentation at code.claude.com. Read the original source in the link below.