Claude Code introduces Projects: A dialog box splits into parallel threads, and the task continues running even when the computer is closed.
On September 17, Anthropic's programming assistant Claude Code officially launched a major upgrade to its Projects feature, transforming AI-assisted development from a single-conversation mode to a multi-threaded automated workflow, marking a substantial step forward for AI programming tools towards autonomous project management.
The core incremental change in this update lies in the execution method: users no longer need to open each conversation individually or manually track each subtask; they only need to describe "what needs to be done," and Claude will break down the task, allocate threads, and execute them in parallel. The official announcement states that these threads will continue working even after the user closes their laptop.
Anthropic stated that the feature is currently available in beta to a select group of Claude Pro and Max subscribers. Access will be gradually expanded over the next few weeks, eventually covering Team and Enterprise plan subscribers.
For developers, this upgrade directly changes how multitasking projects are collaborated on. Workflows that previously required manually switching sessions and piecing together outputs can now be centrally managed by Claude, potentially impacting a wide range of users from individual developers to enterprise engineering teams.
Thread architecture: Claude plays the role of "chief of staff"
The core design of the new version of Projects is a two-layer architecture of "coordinator + worker thread".
Users issue high-level commands in the main project chat interface, and Claude acts as the chief of staff, routing tasks to newly created or existing worker threads; each thread is an independent Claude Code cloud session, performing specific work on its own code branch and repository copy.
Taking a real-world scenario as an example : Users can set the goal to reduce the P75 latency in the application checkout process. Claude will then allocate multiple parallel threads to perform performance analysis on each interface, test optimization solutions, and submit pull requests.
Alternatively, it can connect three code repositories simultaneously: API, Web, and mobile. Claude can create an independent thread for each repository to complete the migration, testing, and PR submission of deprecated interfaces, and inform users which PRs need to be merged first.
If different threads modify the same code, the system will handle the merge conflict in a standard manner, consistent with the regular PR process. Furthermore, a single thread can further invoke sub-delegates, loops, and workflows to break down large tasks and accelerate completion.
Shared memory: Context continues to accumulate across sessions
The new version of Projects is designed for long-cycle or agent-based workflows and supports context accumulation across time and sessions.
All threads share the same memory pool, and Claude continuously accumulates project details over time. For example, it remembers when the release date is changed to Friday, the reason why a certain feature is deprecated, or which colleague needs to be notified before modifying the billing service, thereby reducing the burden on users to repeatedly write prompts.
On a personalization level , Claude also remembers the user's work style and communication preferences. Users can request it to adjust the frequency of progress reports, the pace of creating new threads, and the level of detail in updates.
Alongside the memory function, Projects has also added a "Library" which automatically collects user-uploaded files and various outputs generated by Claude, making it easier to retrieve historical materials and continue to advance based on existing results.
Usage Restrictions and Follow-up Plans
Since a project can run multiple threads simultaneously, and each thread is a complete Claude Code cloud session, the usage of Projects is consumed significantly faster than in single session mode.
Anthropic prompts users to pay attention to project-level usage monitoring and supports selecting models and computing power levels for coordinator chat and worker threads respectively, so as to flexibly manage costs.
Regarding the feature roadmap , Anthropic stated that threads currently only run in the cloud , with features supporting local machine operation, calling local tools and code, and accessing the intranet coming soon.
For the current beta access, it is limited to Pro and Max users who use cloud sessions in Claude Code and do not yet have any projects on the web or desktop.
Existing Projects users can continue to use the existing features normally. Anthropic will complete the upgrade when it is expanded to chat and Cowork features in the future.
Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.