Anthropic has rebuilt the Projects feature in Claude Code so a developer can describe a single goal and watch a coordinator split the work across parallel "threads," each running in its own cloud session. Progress stays visible in the main chat or per thread, including on mobile, and every thread can open pull requests and run tests on its own.
What Changed in Claude Code Projects
The new Projects workflow replaces the previous single-path setup with parallel execution. A user states the goal, a coordinator breaks it into separate threads, and each thread operates independently as a cloud session. That means testing, pull request creation, and code changes no longer have to happen one after another in a single line.
Anthropic also added a shared memory layer that builds across threads over time. A library inside Projects collects every uploaded file and every result, so context accumulates instead of disappearing when one thread finishes.
How Progress Tracking Works
Progress is trackable in two ways. The main chat shows the overall state of the project, while each thread has its own view for granular detail. Mobile tracking is included, so a developer can check thread progress without sitting at the main machine.
Each thread runs tests and opens pull requests directly. That puts quality checks and code review requests inside the same parallel structure rather than requiring a separate pass.
Who Gets Access First
The beta is open to select Pro and Max subscribers who use cloud sessions in Claude Code. Team and Enterprise access will arrive later, and local execution is coming soon. Interested users can join a waitlist.
Where This Fits in Claude Code’s Automation Push
This update is another step toward automated coding inside Claude Code. Anthropic recently made autopilot mode the default in the tool, a change the company said was meant to protect developers from bad approvals. Anthropic claimed autopilot mode outperformed human developers on safety tasks.
More autonomous agents also have a business angle: they are likely to consume more tokens. With parallel threads each running a cloud session, more of the token burn happens under Anthropic’s control rather than the user’s, which shifts some of the throttle from the developer to the company running the infrastructure.
What Developers Can Do with Parallel Threads
- Describe one goal and let a coordinator split it across threads.
- Run tests inside each thread without pausing the others.
- Open pull requests from individual threads.
- Track overall or per-thread progress from the main chat or mobile.
- Build shared memory and a file library that persists across the project.
Why Parallel Session Design Matters
Running each thread as its own cloud session changes how failures are contained. A stalled thread does not stop the rest of the project, and test failures in one area can be addressed while other threads keep moving. The shared library also means files uploaded in one thread are available to the others, reducing repeated setup work.
The waitlist covers the current beta. Team and Enterprise customers will need to wait for the later rollout, and developers who prefer local execution will get that option as the feature expands.
FAQ
What is the new Projects feature in Claude Code?
It is a rebuilt Projects workflow where a user describes a goal and a coordinator splits the work across parallel threads, each running as its own cloud session. Each thread can open pull requests and run tests, and progress is trackable in the main chat or per thread.
Who can access the Claude Code Projects beta?
The beta is open to select Pro and Max subscribers using cloud sessions in Claude Code. Team and Enterprise access comes later, and local execution is coming soon. A waitlist is available.
Does Claude Code Projects work on mobile?
Yes. Progress is trackable in the main chat or per thread, including on mobile.
Related coverage
- Anthropic Releases Claude Opus 4.7 With Extended Reasoning and Agent-Ready Architecture – BizScoreAI
- Anthropic Claude Security Beta Scans and Fixes Code Vulnerabilities – BizScoreAI
- Claude Opus 4.8: Anthropic’s Most Capable AI for Reasoning and Agents – BizScoreAI
This article summarizes reporting from the-decoder.com. See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.