Anthropic launches Claude Code Projects with persistent AI coordinator for developers

Anthropic on Thursday unveiled a redesigned Projects feature for Claude Code, introducing a persistent coordinator agent that can break engineering goals into parallel cloud-based threads, each running as a full coding session on its own branch and repository copy. The threads continue working even after a developer closes their laptop.

From Folder to Conversation

The update, which Anthropic titled "Projects redesigned: from folder to conversation," replaces the earlier project model — essentially a folder of files and a single chat — with an ongoing conversation where Claude acts as both coordinator and dispatcher. Developers describe a goal, and Claude decides what becomes a thread, routes new instructions to existing or new threads, and checks in as requirements evolve.venturebeat+2

Anthropic compared the interaction to briefing a chief of staff: users hand Claude several tasks at once, in any order, and Claude determines where each request belongs. Each worker thread operates on its own git branch and can open pull requests, run tests, and push fixes when CI fails. When two threads modify overlapping code, the result is an ordinary merge conflict resolved like any other pull request.unite+3

Shared Memory and Usage Limits

A shared project memory sits above the individual threads, allowing Claude to retain decisions — such as a release date shifting, a feature being dropped, or which colleague to consult before touching a particular service. A library tab collects uploaded files and artifacts threads produce, giving later work a persistent pool of material.unite+1

Because each thread is a full Claude Code session, projects consume plan limits faster than a single session. Anthropic caps usage at 200 new threads per day across a user's projects, and idle threads can wake and spend again when CI fails or a review comment lands. Users can adjust model and effort settings separately for the coordinator and worker threads.marktechpost

Availability and Competitive Context

The beta launched September 17 for select Claude Pro and Max subscribers using cloud sessions in Claude Code. Anthropic said access will expand over the coming week, with Team and Enterprise plans to follow. Threads currently run only in the cloud; local execution support is coming "very soon," Anthropic said, a detail that matters for enterprises relying on private packages or restricted networks.theverge+2

The release places Anthropic alongside OpenAI, GitHub Microsoft Corporation ↗ 0.52%, and Google Alphabet Inc. ↘ 3.58% in an emerging race around agent orchestration rather than raw coding ability. As The Verge reported, OpenAI's Codex and GitHub's agent tooling already support parallel sessions, but Anthropic is betting that a persistent conversational coordinator — rather than a repository or multi-agent workspace — is the right layer for managing long-running development work.venturebeat+1