News

Tools

CONFIRMED: Claude Code Projects gets a proactive coordinator, shortcuts and read-only links

The November 13 update changes how the Projects coordinator notifies users, adds quick commands and read only sharing, but can drain the plan faster.

Ilustração de um coordenador de IA gerenciando várias tarefas automatizadas em paralelo em telas luminosas

CONFIRMED. Anthropic updated Projects in Claude Code, the feature that organizes long running work into a conversation coordinating several parallel sessions called threads. The official announcement, posted on November 13 by the developer account ClaudeDevs, says the coordinator, the Claude that runs the project's main conversation, now answers questions directly in the project chat and reports on the status of each running thread on its own. The update also adds shortcut commands, scheduled reminders and the ability to share one specific thread as a read only link.

A manager that now speaks up on its own

A project works as a single conversation where the user describes the work and the coordinator decides what becomes a thread, an independent session that usually runs in the cloud with its own copy of the repository and its own branch. Before this update, checking a thread's status meant asking. Anthropic's official documentation, updated alongside the announcement, describes the coordinator answering quick questions without opening a new thread and posting updates when something finishes, gets stuck or needs a decision. In practice, teams already using Projects start getting notices in the main chat instead of opening every thread to check progress.

New commands to keep the thread

The update also brings two shortcut commands inside Projects: slash btw opens a separate thread for a question or side task without pulling the main conversation off topic, and slash fork creates a new thread from the current one to try a different approach without discarding the one already running. A specific thread can also be shared as a read only link with anyone signed in to Claude, as long as that thread never had access to the original user's computer through Remote Control. Scheduled reminders move a thread out of the way until the right time, and the coordinator brings it back into the chat when the deadline arrives.

The limit that comes with the convenience

The official documentation is direct about the cost: Projects can exhaust a plan's usage limit faster than a single session, because each parallel thread draws on the quota independently. Anthropic itself recommends tuning the behavior through the chat, asking the coordinator to run at most two threads at a time, post shorter updates or wait for approval before opening a new thread. Projects remains in public beta, for now limited to Pro, Max and Ultra accounts that already use cloud sessions and don't have legacy projects open; Team and Enterprise accounts don't have access yet, and anyone left out can join the waitlist on Anthropic's site.

Our take

The update doesn't swap the model behind Claude Code, but it changes how the product communicates during long, parallel work, exactly the kind of routine a small team delegates expecting to watch less and trust more. It's worth testing the update frequency controls before leaving several threads running unsupervised, because more autonomy for the coordinator also means decisions made in parallel with less human intervention at each step. Time saved and faster plan consumption need to be weighed together before any productivity promise.

Sources

Anthropic (ClaudeDevs): announcement of the new Projects features; Anthropic: official Projects documentation in Claude Code; Anthropic: Projects redesigned, original September announcement.