Coordinator in Claude Code Projects now reports on its own

Anthropic has changed the coordinator in Claude Code Projects so it posts status updates, pending decisions, finished tasks and blocked work straight into the project chat, and answers questions there too, Alphasignal reported on November 14. If you’re on Pro or Max and run several workers at once, check the effort setting before you fan out: new projects put every worker on Opus at high effort, and each one draws on your regular Claude Code allowance.

The same update adds read-only sharing for single threads, reminders, the /btw and /fork commands inside Projects, and a reworked threads panel. It’s out on the web, the desktop clients and the latest mobile app.

Set the model and effort per role

New projects default to Opus for both roles, with the coordinator on low effort and workers on high. You can change the model and effort level for the coordinator and the worker threads separately. Project usage gets its own line in the usage controls, so you can see what a project costs apart from your other Claude Code sessions.

Projects carry no separate fee, which only moves the bill into the allowance you already have: every worker spends the same Claude Code limits, and several high-effort sessions running at once can use them up faster than a single session. Accounts are capped at 200 new threads per day across all projects. A thread that hits a usage limit pauses, then resumes automatically when capacity comes back.

The coordinator’s habits can be adjusted too. In its own write-up, Anthropic says you can ask Claude to change how often it checks in, how often it starts new threads and how detailed each update is. For small changes the advice is the reverse: work that fits on one branch and in one context window generally costs less in a regular Claude Code session.

What the coordinator now does for you

A project is one long-running conversation with a coordinator that hands work out to threads. Each thread is its own Claude Code cloud session with a separate copy of the repository and its own Git branch, and when two threads touch the same code, the overlap shows up as an ordinary merge conflict once the branches are combined. Threads also share a project memory, where Anthropic’s examples include that a release moved to Friday or whom to check with before touching the billing service.

Until now, finding a thread that was waiting for input meant scanning the thread list yourself. Now pending decisions and blocked work arrive in the main conversation. The threads panel also puts recently opened, updated and pinned threads in a row at the top, and each thread can show thumbnails of the documents, diagrams and other artifacts it produced (you turn the previews on or off in view settings). Project settings and local-folder access were simplified as well.

Parking, branching and sharing threads

Three controls cover work that shouldn’t sit in the active list:

• Reminders hide a thread until a time you set, then put it back in the thread list and notify the coordinator. They’re meant for work that waits on a review, a deployment or an outside reply.
• /btw opens a separate thread for a tangent and keeps the current thread’s goal and context intact.
• /fork starts a new thread from the current thread’s state, for trying another implementation or debugging path.

Both slash commands already worked in regular Claude Code sessions. What’s new is that they now work across coordinated, multi-thread projects.

To share a thread, open its menu, pick Share and send the generated URL. Anyone signed in to Claude can read it, and the rest of the project stays private. Threads that accessed a local computer can’t be shared at all, so shared transcripts can’t leak local file contents, shell output or other data from your machine.

Who can use it

The public beta opened on September 17 for Pro and Max subscribers who joined the waitlist. Team and Enterprise plans are still unsupported. Anthropic let in its entire Pro and Max waitlist in October, and the list is still open to eligible subscribers who don’t have access yet. Don’t confuse this with Claude Projects, the general-purpose workspace for chats and files that Anthropic first introduced in June 2024, VentureBeat notes. The two share a name but do different jobs.

Work that splits cleanly

Projects fit engineering work that breaks into independent tasks and comes back together through branches and pull requests. Anthropic’s examples are performance work, where threads profile endpoints, test optimizations and open pull requests toward a target such as p75 checkout latency, and batch jobs like refactors, dependency upgrades, test repairs or bug triage. Work with strict dependencies doesn’t get that help: you still sequence the pull requests and resolve the conflicts yourself.

Anthropic’s cross-repository migration example shows how much of that ordering the coordinator can hand back. You connect the API, web and mobile repositories, it creates one thread per codebase, and the project reports the order in which the merges need to land.