Anthropic Opens Claude Code Projects Beta

Anthropic has opened an updated version of Projects in Claude Code to select Claude Pro and Max subscribers using cloud sessions. The beta began “starting today” for subscribers who have no existing projects on the web or desktop. Anthropic says access will expand to more Claude Code users on those plans over the coming week, followed by updated Projects across Claude, Team, and Enterprise plans. Pro and Max users without access can join a waitlist; their existing projects continue to work as before and will be upgraded as rollout expands to chat and Cowork.

In a project, users set a goal and choose a repository or other context. Claude proposes work, delegates tasks to parallel threads, coordinates them, reviews their outputs, and assembles results. Users can monitor and steer work in the project chat or individual threads, including from a phone. They can configure the cloud environment, connectors, plugins, instructions, and model. With connected repositories, threads can run tests and open pull requests; with documents, they can read files and draft material. Anthropic’s examples include profiling app endpoints to reduce checkout p75 latency, testing optimizations, and opening pull requests in parallel, as well as migrating callers away from a deprecated v1 endpoint across API, web, and mobile repositories. Claude can identify which pull requests should be merged first.

Each thread is a Claude Code cloud session with its own branch and repository copy. If threads change overlapping code, the conflict must be resolved like any other pull request. Threads can also divide delegated work further through subagents, loops, and workflows. Projects are intended for work spanning multiple replies and parts: threads share memory, so decisions such as a Friday release, a dropped export, or whom to consult before changing billing can inform later tasks. Claude can also remember a user’s working and communication preferences, including how often it checks in, starts threads, and provides updates. A library collects user-added files and Claude-produced artifacts. Projects can run several full Claude Code sessions at once, which can reach usage limits faster; users can check project-specific usage and select models and effort levels for the coordinator chat and worker threads. Threads currently run in the cloud; Anthropic says running them on users’ machines alongside local tools and code and behind their network is coming “very soon.”

Anthropic has opened an updated version of Projects in Claude Code to select Claude Pro and Max subscribers using cloud sessions. The beta began “starting today” for subscribers who have no existing projects on the web or desktop. Anthropic says access will expand to more Claude Code users on those plans over the coming week, followed by updated Projects across Claude, Team, and Enterprise plans. Pro and Max users without access can join a waitlist; their existing projects continue to work as before and will be upgraded as rollout expands to chat and Cowork.

In a project, users set a goal and choose a repository or other context. Claude proposes work, delegates tasks to parallel threads, coordinates them, reviews their outputs, and assembles results. Users can monitor and steer work in the project chat or individual threads, including from a phone. They can configure the cloud environment, connectors, plugins, instructions, and model. With connected repositories, threads can run tests and open pull requests; with documents, they can read files and draft material. Anthropic’s examples include profiling app endpoints to reduce checkout p75 latency, testing optimizations, and opening pull requests in parallel, as well as migrating callers away from a deprecated v1 endpoint across API, web, and mobile repositories. Claude can identify which pull requests should be merged first.

Each thread is a Claude Code cloud session with its own branch and repository copy. If threads change overlapping code, the conflict must be resolved like any other pull request. Threads can also divide delegated work further through subagents, loops, and workflows. Projects are intended for work spanning multiple replies and parts: threads share memory, so decisions such as a Friday release, a dropped export, or whom to consult before changing billing can inform later tasks. Claude can also remember a user’s working and communication preferences, including how often it checks in, starts threads, and provides updates. A library collects user-added files and Claude-produced artifacts. Projects can run several full Claude Code sessions at once, which can reach usage limits faster; users can check project-specific usage and select models and effort levels for the coordinator chat and worker threads. Threads currently run in the cloud; Anthropic says running them on users’ machines alongside local tools and code and behind their network is coming “very soon.”