Claude's redesigned Projects feature is rolling out to select Pro and Max users first. Here's who qualifies and what's excluded for now.
What is the new Claude Projects feature? #
Claude Projects has been redesigned around a coordinator model. Instead of running isolated chats inside a folder of shared context, a project is now one long-running conversation that delegates work to separate threads. Each thread runs as a full Claude Code cloud session on its own branch, can split into its own sub-agents, and keeps working even after you close your laptop. The coordinator briefs the work, threads execute it, and everything shares a pool of project memory, files, and instructions.
Access to this redesigned version is limited during rollout, and that limitation is the practical thing to understand before you go looking for it in your account.
TL;DR #
- Access is currently restricted to selected Pro and Max subscribers using Claude Code cloud sessions, and even within that group the rollout is staged, so not everyone on those plans has it yet.
- Existing web and desktop project users are excluded for now , meaning if you already have Projects set up in Claude chat or Claude Cowork, you may not be first in line for the new coordinator version.
- Team and enterprise access is coming later , along with permission controls that let organizations set view or edit rights on shared project context by email or org-wide.
- Bedrock, Vertex, and Foundry deployments don’t support this yet , so anyone running Claude through those platforms is excluded from both the new Projects coordinator and the related Agents.md support.
- Local file access isn’t part of this either , since threads run in the cloud only, meaning the coordinator can’t see local files or local tools the way a locally-running agent could.
- Parallel threads consume usage allowance faster , since each thread is its own cloud session, so running several at once burns through your plan’s limits quicker than a single chat would.
- A retrieval layer (project knowledge search) works in the background with no setup required, pulling relevant files into a thread instead of the entire project context every time.
Which plans get access to the new Projects? #
The redesigned Projects beta is available to selected Pro and Max subscribers who are using Claude Code cloud sessions. That’s a narrower group than “everyone on Pro or Max.” The rollout specifically excludes users who already have Projects set up on web or desktop, at least for this initial phase. So if you’re a longtime Claude chat user with an established Projects folder, you’re not automatically in the first wave.
Team and enterprise plans are not included yet. Anthropic has said that access for those tiers, along with the permission-level controls that come with them, is coming later. Local execution (running threads against local files and tools rather than purely in the cloud) is also slated for a future release, not the current beta.
In short: Pro and Max get first access, and only a subset of those users, gated by whether they’re on Claude Code cloud sessions and don’t already have legacy Projects running elsewhere.
What’s excluded from the current rollout? #
Three exclusions matter most for anyone planning around this feature.
First, platform support. At launch, this doesn’t work on Bedrock, Vertex, or Foundry. Anthropic has stated explicitly that support for those deployment platforms isn’t available yet. If your organization runs Claude through AWS Bedrock or Google Vertex, you’re not getting the new Projects coordinator or the related Agents.md instruction-file support through those channels right now.
Second, local files and local tools. Because threads run entirely in the cloud, the coordinator can’t see anything on your machine. That also means the main conversation can’t touch connectors like MCPs directly. Those integrations are handled independently inside individual threads rather than at the coordinator level.
Third, existing web and desktop Projects users are carved out of the initial rollout. Anthropic has said the redesigned experience will eventually come to Claude chat and Cowork as well as Claude Code, where it currently lives, but that expansion hasn’t happened yet.
How does the project coordinator actually work? #
You brief the top-level conversation the way you’d brief a team lead, not the way you’d type a command into a terminal. The coordinator decides whether your request needs a new thread or belongs in one that’s already running. Each thread that spins up is a complete Claude Code cloud session with its own branch and its own copy of the relevant repository, and it can further split into sub-agents and workflows underneath it.
Every thread gets automatically populated with the project’s repos, uploaded files, project instructions, and a memory file that carries decisions across conversations. That’s a meaningful shift from the old setup, where individual chats inside a project shared reference material but had no way to communicate with each other. If you wanted one chat to know what happened in another, you had to manually point it there.
A library tab collects both the files you upload and the files Claude produces, so outputs don’t get scattered across a dozen disconnected sessions.
#
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
Behind the scenes, projects now have a retrieval layer, effectively a form of retrieval-augmented generation scoped to project content. Instead of every file into every conversation and running into context limits, Claude gets a project knowledge search tool that pulls in only what’s relevant to the question at hand. There’s no setup required on your end. As a project’s total content shrinks back under the context limit, it reverts to everything directly again.
Is Claude Projects worth using right now? #
That depends heavily on which plan you’re on and what you need it for. If you’re not yet in the eligible group of Pro or Max subscribers on Claude Code cloud sessions, the question is moot for now, since the feature simply isn’t available to you yet.
For those who do have access, the appeal is real but the limitations are worth weighing. The coordinator model solves a genuine pain point: previously isolated chats now share memory and context automatically, and the retrieval layer should reduce how often you hit context walls and have to start over. Running work in the cloud means tasks can continue after you close your laptop, which is useful for longer jobs. The tradeoffs are that it’s cloud-only (no local file or tool access yet), it doesn’t support MCP connectors at the coordinator level, and running multiple threads in parallel eats into your usage allowance faster than a single conversation would. It’s also new enough that Anthropic itself frames it as an early beta rather than a finished product. Teams that need permission controls, shared editing rights, or enterprise-grade access controls will have to wait for that tier to roll out.
Frequently Asked Questions #
Which Claude plans currently support the new Projects coordinator?
Selected Pro and Max subscribers using Claude Code cloud sessions. It’s not available to everyone on those plans simultaneously, and it excludes users who already have Projects configured on web or desktop.
Does Claude Projects work with Bedrock or Vertex?
No. Anthropic has stated that support for Bedrock, Vertex, and Foundry isn’t available at launch for this feature or for the related Agents.md instruction file support.
Can Claude Projects access local files?
Not currently. Threads run entirely in the cloud, so the coordinator has no access to local files or local tools. Local execution is planned for a future release.
When will Team and Enterprise plans get access?
Anthropic has said Team and Enterprise access, including permission controls for sharing project context by email or organization-wide, is coming later. No specific date has been confirmed.
Does running multiple threads cost more?
Each thread is its own full Claude Code cloud session, so running several in parallel draws down your plan’s usage allowance faster than working in a single conversation.