Anthropic's Claude Code introduces Projects feature
Anthropic's Claude Code launched Projects, allowing a single conversation to spawn multiple independent cloud threads that run in parallel, even with the computer closed. Each thread has its own context, sandbox, and git branch. Features include shared memory, PR tracking, and scheduled routines.
Key insights
Projects enables parallel, persistent coding threads from one conversation
Claude Code's Projects feature lets users assign tasks in one conversation, and Claude decides whether to answer directly, spawn a new thread, or delegate to an existing thread. Each thread is an independent cloud Claude Code session with its own context window, code copy, and git branch. Threads continue running even after the laptop is closed.
Shared background and memory across threads
Each new thread starts with the same project instructions, repository, and project memory. Decisions that Claude is told to remember are available to all subsequent threads. An overview panel groups threads by status, with threads waiting for user input shown in a 'Waiting on you' section with desktop notifications.
PR tracking and scheduled routines automate maintenance
Threads that open PRs automatically monitor them: if CI fails or review comments arrive, the thread pushes fixes and notifies the user when checks pass. Users can also set up routines that run automatically on a schedule within the project. The feature works without a code repository for research and report writing.