Anthropic has rebuilt Claude Code’s Projects feature so a single user can run a team of AI agents working together on one task, with shared memory, goals, and a library of files. The company announced the relaunch today, calling it a major step toward managing multiple agents in a coordinated way.
The new setup puts Claude Code agents in a similar position to Grok Bot and other tools that handle groups of AI agents. Each project has threads running separate tasks at the same time, with a coordinator agent managing the whole operation. That means a single project can hold several agents working on different parts of the same job, passing information back and forth as they go.
Threads, Branches, and Merges
Under the hood, each thread runs as its own Claude Code cloud session, working on its own branch of the repository. The coordinator keeps everything organized, but if any threads work on the same code at the same time, the overlap gets resolved as a merge conflict — the same kind of conflict that shows up in any other pull request.
Threads can also split their assigned work into pieces using subagents, loops, and workflows when needed, so large jobs finish faster. Users can watch each thread individually, or keep an eye on the whole project through the main chat window.
How Threads Talk to Each Other
The coordination model is the heart of the feature. Threads do not work in isolation — they share a common memory and a common goal, and the coordinator directs the flow of work between them. When one thread produces a file or a piece of data, another thread can pick it up and use it, passing the work along through the group.
That design removes the need to move files between sessions manually. Instead, the system handles the handoff automatically, letting agents build on each other’s work without interruption.
Access and Availability
The updated Projects feature is available in beta starting today for “select Claude Pro and Max subscribers.” Access expands to all Pro, Max, Team, and Enterprise users later, as well as Cowork and regular Claude chats.
Threads currently run in the cloud, but Anthropic says support for local tools and code is coming “very soon.” That suggests Anthropic is moving toward supporting local execution, though the company has not detailed how that will work.
What This Changes
The key shift is scale. Before, a user managed one agent at a time. Now, a user manages a team of agents working together on a single task. The coordinator handles the logistics — resolving conflicts, and keeping the group pointed toward the goal.
The ability to split work across subagents, loops, and workflows means large assignments can finish faster, since the main thread does not have to do everything itself. That is a practical benefit for users who need to move through a lot of code quickly.
How It Compares to Grok Bot
| Feature | Grok Bot | Claude Code Projects |
|---|---|---|
| Model | Group management tool | Agent coordination |
| Structure | Threads running separate tasks | Threads running separate tasks |
| Coordinator | Yes | Yes |
| Merge handling | Not specified | Merge conflicts resolved like PRs |
| Subagents | Not specified | Yes |
Why Multiple Agents Matter
Running multiple agents is useful for several reasons. A team of agents can divide a large project into smaller pieces, with each thread taking on a different part of the job. That parallel approach speeds up development, especially on large projects.
Shared memory and a shared library also mean agents can refer to the same files and artifacts without duplicating work.
The Coordinator’s Role
The coordinator is the central figure in the system. It resolves merge conflicts, and monitors the overall progress of the project. Without it, the threads would be independent agents working in parallel but not coordinating their efforts.
That role is what sets Claude Code’s Projects apart from simple parallel processing. It is not just running multiple agents at once — it is running them as a team, with a clear division of labor and a single point of direction.
What Comes Next
The beta is live now for select subscribers, and Anthropic says the full rollout to all Pro, Max, Team, and Enterprise users is coming later. Local support is on the way, too, though Anthropic has not given a timeline for that beyond “very soon.”
The company has not detailed the full feature set for the expanded release. What is known is that the core coordination model — threads, a coordinator, shared memory, and merge resolution — will be available to all tiers of users once the rollout completes.
“Very soon”
Our View
This is a significant move for Claude Code. Managing multiple agents as a coordinated team is a natural step, and Anthropic’s implementation offers a clean model for how it could work.
The merge-resolution approach is clever — treating agent overlap like a pull request means developers already know how to handle the conflicts. And the subagent, loop, and workflow tools give users real flexibility in how they split their work.
There is room for improvement. Local support is coming “very soon,” but no date has been attached to it. And while Anthropic has not detailed the full feature set for the expanded rollout, the beta is limited to select subscribers.
Still, the direction is clear. Anthropic is building toward a more capable version of Claude Code, one where multiple agents can work together on a single task with coordination and shared memory. The company has already shown what that looks like — now it just needs to deliver it to everyone.
See the a run of 29 images at The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.