Anthropic Opens Claude Code Projects to All Pro Users: Cloud Parallel Sessions and Auto-PR Loops

Claude Code Projects Now Fully Open to Pro and Max Users on the Waitlist

Anthropic has granted full access to Claude Code Projects for all Pro and Max users previously on the waitlist. The feature is currently in public beta and being rolled out gradually; it is not yet available for Team and Enterprise plans.

I’ve been using Claude Code Projects extensively lately, and I think it’s a well-designed feature. It allows you to act like a personal assistant by continuously dispatching tasks in the main session, while also providing visibility into the execution of each subtask via Threads. This maintains a record and facilitates follow-ups.

I recommend watching this video—it explains things very clearly.

In a previous post, I explained the differences between Codex Projects and Claude Projects:

If you remember old-school forums, a Codex Project is like a forum board—a container or category where all content related to that board/project lives. You can start new sessions within it repeatedly.

Claude Projects (the new ones) are like Slack Channels. They are also categories or containers, but all messages flow together. You can throw anything into them, but if you want to dive deep into a specific subtopic, you open a Thread and continue the discussion there.

Forums tend to go off-topic easily; discussions drift from the original subject, contexts get messy, and only recent replies remain clear.

Slack is simple and convenient. You don’t need to worry about which topic a message belongs to—just send it to the Channel. A bit of context clutter is acceptable because you can use Threads to focus.

Projects manage the coordination of "running multiple Claude Code sessions simultaneously." Previously, if you wanted to work on several tasks in parallel, you had to assign tasks to each session manually, re-explain the background every time, and check back individually to see who was done. Now, you interact with a single "Project Conversation," where Claude acts as the coordinator, breaking down goals into smaller tasks and spawning a "Thread" for each one.

Each thread is a complete Claude Code session, running by default in the cloud on its own Git branch. Even if you close your laptop, the work continues, and you can monitor progress from your phone. If a task requires local resources like a database, simulator, or internal network interface, you can switch that thread to run on your own machine (via Remote Control, provided your computer is on).

Coding threads automatically create Pull Requests (PRs) after making changes and then keep monitoring them: they submit fixes if automated tests fail and address review comments when reviewers leave feedback. An official demo example involves troubleshooting website performance issues: multiple threads fix different parts, generate a before-and-after performance comparison report, and set up a daily routine check. If performance regression is detected, it automatically spawns a thread to investigate, waiting for you to handle it upon your next visit.

Projects are best suited for work that doesn’t fit into a single session and generates continuous new tasks, such as migrating an app away from a deprecated dependency or synchronizing changes to the same API across web, mobile, and backend repositories. You can also use them without touching code—for example, uploading batches of contracts or customer support tickets and asking questions iteratively.

All threads share the project instructions and project memory. If you say once, "Always create PRs from the main branch," every subsequent thread will follow that rule.

[Usage Warning]

Since each thread is a full session and multiple can run concurrently, projects consume plan quotas faster than single sessions. Anthropic specifically warns that Pro users may hit their limits sooner. New projects default to using Opus, and the reasoning effort (how much thinking the model invests at each step) defaults to High. The Claude Code team recommends: Keep the coordinator at the default low effort, as its primary role is task dispatching—increasing effort offers little benefit. For threads, consider switching to a smaller model by default, upgrading to a larger model only for complex tasks. You can also instruct Claude to limit the maximum number of concurrent threads or ask it to show you the plan before executing.

If a thread hits the quota limit, it will pause and automatically resume once the quota resets. Unmanaged tasks will consume quota from your next billing period. To avoid this, you can pause the entire project in the settings.

[How to Use]

At

↳ Quote @ClaudeDevs (ClaudeDevs): We just let in every Pro and Max user from the Claude Code Projects waitlist!

If you're new to Claude Code Projects, here's a 4 minute walkthrough to get you started: