Claude Code Projects Relaunch as Multi-Agent Cloud Threads, Each on Its Own Branch
- What happened
- Anthropic relaunched Projects in Claude Code on 17 September 2026 as a coordinator directing parallel cloud threads, each on its own branch and repo copy.
- Why it matters
- It turns Claude Code into a multi-agent workflow with merge conflicts as the collision model, but parallel threads hit usage limits faster and access is a Pro/Max beta.
- What to do
- If you get access, start with a few threads and watch your usage before fanning out; Claude Code stays conditional on cost control.
Our verdict: a real step toward multi-agent coding in Claude Code, but it does not change our call. Claude Code stays conditional in our directory, and the condition is cost control. Projects makes that condition matter more, not less: Anthropic itself says parallel threads can reach usage limits faster. We have not tested the new Projects yet; it is a beta for select Pro and Max subscribers.
What happened
On 17 September 2026, Anthropic redesigned Projects in Claude Code. A project is now a conversation with a coordinator that directs work across parallel threads.
- The coordinator scopes the request, delegates the work, coordinates the parallel threads, reviews the outputs and assembles the finished result.
- Each thread is a Claude Code cloud session working on its own branch and its own copy of the repository.
- Collisions surface as ordinary merge conflicts when threads touch the same code, the same model as a pull-request workflow.
- Shared project memory carries project details, decisions and communication-style preferences across threads.
Who gets it, and when
| Stage | Who |
|---|---|
| Now | Beta for select Pro and Max subscribers using cloud sessions in Claude Code (waitlist for the rest) |
| Over the coming week | More Claude Code users on Pro and Max |
| Later | Team and Enterprise |
Existing Pro and Max projects keep working as they do today; Anthropic says it will upgrade them as the rollout expands to chat and Cowork.
Threads run only in the cloud for now. Anthropic says running them "on your machine alongside your local tools and code and behind your network is coming very soon".
Why it matters
- Claude Code moves from one agent to an orchestrated team. Branch-per-thread plus merge conflicts is a familiar, auditable collision model rather than agents silently overwriting each other.
- Spend scales with fan-out. Every thread is a full Claude Code session. A coordinator running several threads in parallel burns through Pro and Max usage limits faster than one session doing the same work in sequence, and Anthropic's own announcement flags this.
- Cloud-only is a constraint today. Teams whose code cannot leave their network must wait for the promised local threads.
What changes for you
- If you have beta access: start with two or three threads on a well-scoped task and watch your usage before fanning out wider.
- If you are on Pro or Max without access: join the waitlist; the rollout is expected over the coming week.
- If you are on Team or Enterprise: nothing changes yet. Your rollout comes after Pro and Max.
- If your repo must stay on-premises: wait for local threads; there is no date beyond "very soon".
FAQ
Do my existing Claude Projects break? No. Existing Pro and Max projects keep working as they do today and will be upgraded as the rollout expands.
Can threads run locally? Not yet. Threads run in the cloud today; local execution is "coming very soon" with no firm date.
Does this change the Claude Code verdict? No. Claude Code remains conditional on cost control, and parallel threads raise the stakes of that condition.
What to do
- 1 If you have beta access, start with two or three threads and monitor usage before fanning out.
- 2 Pro and Max users without access can join the waitlist; rollout is expected over the coming week.
- 3 Teams that need on-network execution should wait for local threads.
Affected tools & models
Never need to catch up again
The weekly delta — only verdict changes and act-now items. No digest filler.