What are the new Claude Code Projects?
Anthropic is introducing a new version of Projects that changes the feature from a collection of separate chats into a more agentic workspace.
In the new beta, Anthropic says:
A Project is one conversation. You describe what you need as the work evolves, and Claude can divide that work into parallel cloud threads.
Claude reports which pieces are complete and which are waiting for input. The Project keeps shared context around the work, so the user does not have to manually rebuild the same repository and instruction context in every thread.
The first rollout is inside Claude Code, not a universal replacement of all existing Projects at once.
Official source: Claude Help Center — What are projects?.
How parallel threads work
Parallel threads let Claude work on more than one part of a larger objective at the same time.
For example, one coding Project could have separate threads working on:
- reading the repository and mapping the architecture
- implementing a backend change
- updating frontend code
- running tests or investigating failures
- drafting documentation
The important difference is that these are not simply five unrelated Claude chats that the user opened manually. Anthropic says the threads belong to the same Project and begin from shared Project context.
This makes the new Projects experience closer to orchestrated parallel work: one top-level Project conversation can fan work out into several cloud workstreams and then surface their status back to the user.
Are Claude Code Projects using “parallel agents”?
The search phrase makes sense, but Anthropic’s official Projects documentation uses the term parallel threads.
That distinction matters because Claude Code already has separate concepts called subagents, and Anthropic also documents running multiple independent Claude Code sessions in git worktrees.
So the clean terminology is:
| Term | What it means |
|---|---|
| Project parallel threads | Cloud workstreams Claude creates inside one new-version Project. |
| Subagents | Specialized agents that a Claude Code session can delegate tasks to. |
| Parallel sessions | Separate Claude Code sessions the user runs at the same time. |
| Worktrees | Git-isolated working directories that keep parallel code changes from colliding. |
In other words, “parallel agents” is a useful informal description, but it is not the exact product term Anthropic uses for the new Projects beta.
Every thread starts with the Project’s shared context
Anthropic says each thread starts with the Project’s:
- files
- repositories
- instructions
- memory
This is one of the main reasons the feature is more useful than opening several independent Claude Code chats.
If a Project contains the repository, project instructions and accumulated memory, a new thread does not have to be taught the basic context again from scratch.
That shared foundation also gives Claude a more coherent way to coordinate work across several tasks that belong to the same larger objective.
The Project Library collects inputs and outputs
The new Projects experience includes a Library.
Anthropic says the Library collects:
- files the user adds
- files Claude produces
This gives the Project a persistent place for working material instead of treating every generated artifact as an isolated attachment inside a single thread.
For a coding workflow, that can include specifications, reference files, generated documentation and other artifacts Claude creates while the Project is running.
Do Claude Code Project threads keep running if you close your laptop?
Yes. Anthropic says the new Project threads run in the cloud and keep going if you close your laptop.
This is one of the most important differences from a purely local terminal workflow. The Project can continue cloud work while the user is away from the machine.
That does not mean every local-only dependency or permission magically remains available. A thread still depends on the environment and resources available to that cloud execution context. But the Project itself is not tied to keeping the laptop awake for the thread to continue.
Who can access the new Projects beta?
Anthropic says the new version is currently available in beta to selected Pro and Max subscribers who use Claude Code.
If your account has access, you should see Projects in:
- the sidebar at
claude.ai/code - the Code tab of the Claude Desktop app
Anthropic says more Pro and Max Claude Code users will receive access as the staged rollout expands.
Pro and Max users who do not have access yet can join the waitlist.
Current beta access: Anthropic Projects Help Center.
Why Claude Code Projects may not be showing
If you are on Pro or Max and cannot see the new Projects interface, that does not necessarily mean something is wrong with your account.
The beta is being rolled out in stages. Work through these checks:
- Confirm you are on Pro or Max. The current beta is limited to selected users on those plans.
- Open
claude.ai/code. The new Projects entry should appear in the Code sidebar for accounts that have access. - Check the Code tab in Claude Desktop.
- Make sure you are signed into the same Claude account that has your Pro or Max subscription.
- If Projects is still absent, your account may simply not have reached the staged beta yet.
- Join the waitlist if Anthropic presents that option for your account.
Current answer: Pro or Max is necessary for this first Claude Code beta, but it does not mean every Pro or Max subscriber receives it immediately.
Old Claude Projects vs the new Projects beta
This is another source of confusion because Anthropic is using the same product name during a transition.
| Area | Existing Projects | New Projects beta |
|---|---|---|
| Availability | Available to all users, including Free. Free users can create up to five Projects. | Currently selected Pro and Max Claude Code users. |
| Structure | Self-contained workspace with chat history and project knowledge. | One ongoing Project conversation that can branch work into parallel cloud threads. |
| Parallel work | Not the defining interaction model. | Core part of the new design. |
| Background cloud threads | Not described as the current Project model. | Threads can continue when the laptop is closed. |
| Shared thread starting context | Chats use Project knowledge/context. | Every thread starts with Project files, repositories, instructions and memory. |
| Rollout status | Existing product remains available. | Beta, rolling out in stages. |
Anthropic says existing Projects in chat and Cowork continue to work as they do today. On Pro and Max, those Projects will be upgraded as the new version expands to chat and Cowork.
How Claude Code Projects affect usage limits
There is no separate unlimited Project allowance.
Anthropic states directly that Projects draw from your plan’s usage like any other use of Claude, and that running several threads at once uses it faster.
This means parallelism trades time for consumption. If three or four threads are reading repositories, reasoning and producing output simultaneously, your plan can make more progress in wall-clock time, but it can also consume your included allowance more quickly than one sequential conversation.
Claude Code activity is also part of the same broader subscription usage system. Anthropic says Pro and Max usage is shared across Claude and Claude Code, so ordinary Claude conversations and coding work contribute to the same plan limits.
Projects usage: Projects Help Center. Shared Pro/Max usage: Use Claude Code with your Pro or Max plan.
Pro vs Max: how much more room does Max give parallel Projects?
Anthropic offers Max in two individual tiers:
| Plan | Included session capacity relative to Pro |
|---|---|
| Pro | Baseline |
| Max 5x | 5× Pro usage per session |
| Max 20x | 20× Pro usage per session |
Anthropic says Max session-based limits reset every five hours. Max also has weekly limits across models, with the reset time assigned to the account.
Paid Claude plans expose usage progress in Settings → Usage, including the five-hour session window and weekly usage progress where applicable.
Do not interpret “Max 20x” as 20 Project threads. The 20x figure describes plan usage capacity relative to Pro, not a published thread-count entitlement.
Max limits: What is the Max plan?. Usage dashboard: Usage limit best practices.
How many parallel Project threads can you run?
Anthropic’s current Projects Help Center page does not publish a fixed numeric limit such as “Pro gets three threads” or “Max gets ten threads.”
The documented constraint is usage: several threads running simultaneously draw from the same plan allowance and therefore use it faster.
This is an important distinction because it prevents two common misconceptions:
- Max 5x does not mean exactly five concurrent Project threads.
- Max 20x does not mean exactly twenty concurrent Project threads.
Anthropic can also impose additional capacity or feature limits, so the live product UI and account-level usage indicators are the safest authority for an individual account.
Claude Code Projects vs git worktrees
Projects do not make git worktrees obsolete.
Anthropic’s Claude Code team recommends running multiple sessions in separate worktrees when several coding tasks need filesystem and branch isolation. Claude Code has native worktree support:
claude --worktree
Worktrees solve a different problem from Projects:
| Feature | Best understood as |
|---|---|
| New Project threads | Claude-orchestrated parallel cloud work inside one Project context. |
| Git worktrees | Separate working copies/branches so parallel code changes do not collide. |
For large code changes, the strongest workflow may use both concepts: one Project coordinates the objective while isolated worktrees keep independent implementation branches from stepping on each other.
Worktree guidance: Claude Code power user tips.
Claude Code Projects vs subagents
Claude Code subagents are specialized delegates inside the Claude Code agent system. Anthropic describes them as useful for parallel work such as searching different parts of a codebase, reviewing a diff from different perspectives or generating competing implementations.
Subagents can also be configured with their own:
- system prompt
- model
- tool permissions
- color/name
- worktree isolation
The new Projects beta is a higher-level workspace concept. It gives one Project conversation shared files, repositories, instructions and memory, and Claude can split that Project’s work into parallel cloud threads.
So a useful mental model is:
Project = the persistent objective and shared workspace → threads = parallel Project workstreams → subagents/worktrees = lower-level Claude Code techniques that can also help execute isolated tasks.
Subagent guidance: Claude Code user FAQ and Claude Code power user tips.
What about Team and Enterprise?
Team and Enterprise customers can already use Claude Code according to their plan/seat configuration, but that is separate from access to this specific new Projects beta.
Anthropic’s Projects page says the staged rollout currently begins with selected Pro and Max Claude Code users, and that chat, Cowork, Team and Enterprise will follow later.
That means a Team or Enterprise seat having Claude Code access does not automatically mean the new Projects interface has already been enabled for that organization.
Claude Code on organizations: Team and Enterprise Claude Code access. Projects rollout: Projects Help Center.
How to use parallel Projects without wasting your limits
Parallel work is most useful when the tasks can genuinely proceed independently. Spawning several threads for tightly coupled work can consume extra usage without saving much time.
- Give the Project clear shared instructions. Put repository conventions, acceptance criteria and important constraints in the shared Project context.
- Split independent workstreams. Good candidates include research, implementation, tests and documentation.
- Avoid duplicating the same investigation across several threads unless you intentionally want independent answers.
- Use worktrees when parallel code edits need branch isolation.
- Use subagents when a specific task benefits from a specialized role or tool set.
- Watch Settings → Usage. Anthropic exposes your session and weekly consumption there on supported paid plans.
- Reduce model effort for routine work if appropriate. Anthropic notes that higher effort levels consume more usage.
- Reserve large parallel bursts for work where finishing sooner is worth the faster allowance consumption.
BeeVaults takeaway: the new Claude Code Projects beta turns one Project into an orchestrated cloud workspace where Claude can create parallel threads that share files, repositories, instructions and memory. It is currently a staged beta for selected Pro and Max Claude Code users, not a universal replacement for existing Projects yet. The convenience comes with a straightforward tradeoff: every parallel thread draws from the same plan usage pool, so more simultaneous work can reach Pro or Max limits faster.