Anthropic tests team collaboration Projects feature in Claude Code

The core of this update is the ability to spawn multiple threads within a single project session. Previously, one project was tied to one conversation; now, different workflows within the same project can be split off into separate threads.

A collaboration model resembling Slack

This structure overlaps with Slack's concept of channels and threads. The project session itself becomes a channel, with multiple people and Claude staying present in it simultaneously, while individual tasks branch off into threads. Anthropic has already been running Claude Tag on Slack, referencing codebases tied to channel and thread context — and this new collaborative Projects feature appears to bring that same flow directly into the Claude app. Even without Slack, an entire team can now share a single workspace inside Claude.

There are also signs Anthropic has been investing in Claude Tag for Slack. In a post on X, Noah Zweben said of the Claude Tag improvements, "unprompted messages are down about 45%." The comment suggests Claude has gotten better at judging when to speak up and when to stay quiet — a separate feature from this collaborative Projects update, but one that indicates Anthropic has been focused on context-awareness capabilities around the same period.

Element	Existing Managed Projects	This Update
Repository connection	Attached temporarily per session as needed	Pinned as permanent context at project creation
Session structure	Single conversation per project	Multiple threads spawned within a project
Context management	Managed manually by the user	Periodically reorganized by Claude itself

Claude maintains the project on its own

Another notable detail is that Claude reportedly maintains this project by itself. Claude is said to periodically review, reorganize, and refine context files and memory. This turns a project from a static conversation record that's saved once and stays frozen into something closer to a living workspace that's continuously groomed. Anthropic reportedly sees this collaboration model as moving beyond traditional pair programming or solo work with AI, toward multiple developers and Claude working together on a single project. Internally, this approach is said to have already been in use for some time.

Still limited access, for now

Rollout appears to be expanding beyond friends-and-family testing or an early alpha stage. Some early-access organizations already appear to have access, but no confirmed public rollout schedule has been announced. As such, there's currently no confirmed menu path or step-by-step usage guide for general users to actually try it out. Still, the direction is clear — Anthropic is pushing Claude from a coding assistant that helps a single developer toward organizational infrastructure used by multiple people together.

Editor's take

What makes this feature interesting isn't what was newly built, but what was removed. Until now, AI coding tools have essentially been designed for single-user use. One developer would throw a prompt and get an answer, while team collaboration was handled by tools outside the AI — like Slack or GitHub comments. This update erases that boundary. The moment a repository is permanently attached to a project and multiple people enter the same session split across threads, Claude stops being a personal assistant and becomes a participant in a team channel.

The shift feels sharper when compared generation to generation. Just months ago, Claude Code was focused on adding "single-agent management" features like setting budget caps and consulting an advisor model. Now it has pivoted toward making the session itself something multiple people share. Set alongside Cursor testing a similar direction with its codebase and review tabs in private beta, it's fair to read this as the AI coding tool market as a whole shifting its center of gravity from "personal assistant" to "team workspace."

For Korean development teams, it's too early to factor this feature into adoption plans just yet — it's only open to early-access organizations, and there's no public release schedule. Still, the underlying design — pinning a repository as context and splitting sessions into threads — is reason enough to review internal team workflows right now. If a team is currently re-explaining repository context every time in whatever AI coding tool it uses, that repetitive work represents a cost that's about to disappear.

In the coming weeks, Anthropic is likely to expand this feature's rollout to the organizational level, or formally announce it integrated with Claude Tag on Slack. With competing tools like Cursor and GitHub testing repository integration and review features around the same time, the next axis of competition among AI coding tools looks set to shift from model performance to how well a team can be made to work together in the same space.