Anthropic relaunches Claude Code Projects in beta to coordinate multiple parallel cloud sessions with shared memory, goals, and files
Feature update ● Confirmed 86% confidence first seen
Anthropic has relaunched Claude Code Projects, allowing a coordinator to split an engineering goal into multiple parallel cloud sessions that run on separate repository branches. The results are assembled for review, overlapping changes are handled via standard merge conflicts/PR-style workflows, and threads are designed to continue running after a user closes their laptop. Coverage also notes that parallel threads increase token/usage consumption and can reach plan limits sooner during the beta.
Decision brief
- What changed
- Anthropic relaunched Claude Code Projects in beta, adding a coordinator that can split one engineering goal into multiple parallel cloud sessions with shared memory, goals, files, and separate repository branches. The sessions can keep running in the cloud after a user closes their laptop, and overlapping edits are resolved through standard PR and merge-conflict workflows.
- Why it matters
- This changes how engineering teams can use coding agents operationally: instead of one sequential assistant session, teams can coordinate several concurrent branches of work inside one project and review the assembled output through familiar software-development workflows. For decision-makers, the tradeoff is clear in the coverage: potential productivity gains from parallel execution come with materially higher usage consumption because each thread counts as a full Claude Code session, which can push teams into plan limits faster during the beta.
- Evidence
- All three cited outlets describe the same core product change: a coordinator launching parallel Claude Code cloud sessions on separate repo branches with shared project context. The Verge and MarkTechPost align on branch-based parallel work and merge-conflict handling, while The New Stack independently adds the cost/usage implication that each thread is billed as a full session and may exhaust plan limits sooner in beta.
- What remains uncertain
- The coverage does not quantify actual productivity gains, defect rates, or net cost per completed engineering task, so ROI remains unverified. It is also unclear how broadly beta access will expand, what additional pricing tiers will look like, and how reliably long-running background sessions perform across different team sizes and codebase complexities.
- Monitor next
- Watch for Anthropic to publish broader beta access, pricing or usage policy details, and any team-level metrics on throughput or cost per project when using parallel threads.
Analytical support, not advice — assumptions and open questions stated above.