Anthropic Reinvents Claude Projects: One Conversation Orchestrating Parallel Cloud Threads

On September 18, Anthropic officially announced a major redesign of Claude Projects, rolling out Projects (beta, gradually to some Pro/Max users) in the Claude Code desktop and web apps. The core shift: projects go from "folders" to "one ongoing conversation"—users toss in requirements in any order within the same session, and Claude automatically breaks down tasks, dispatches parallel threads, coordinates the work, and consolidates results. This is a directional overhaul of Anthropic's project workflow, and it's worth watching whether it can meaningfully improve collaboration efficiency for multi-task engineering.

Confirmed

• Every thread is a cloud-hosted Claude Code session that keeps running even when the local machine is shut down or offline, and users can jump in anytime from their phone. Officially, access to local files, tools, and intranet resources is "coming soon"—for now, threads cannot access the local environment.
• Users can hand off multiple tasks at once, like briefing a chief of staff; Claude automatically routes them to new or existing threads and proactively asks for confirmation when decisions are needed, with an Overview panel collecting all pending decision points.
• Projects have a shared-memory mechanism: each thread writes to and reads from shared memory, so Claude remembers facts like "the launch moved to Friday" as well as conventions like "ask so-and-so before touching billing services." A project's library stores both user-uploaded files and content generated by Claude.
• Details from the official blog demo show Claude can split parallel tasks, open PRs, collaborate across repos, and review output from each thread.

Why it matters

• This upgrades AI from "one chat does one thing" to "one chat commands multiple threads working in parallel"—an orchestration model that sharply cuts the cost of juggling multiple sessions.
• Continuous cloud execution plus mobile access means long-running tasks no longer depend on a local machine staying online, changing the spatiotemporal constraints of dev workflows.
• Shared memory keeps project context consistent across threads, a key step toward long-running, collaborative AI engineering agents.

2026-09-18 ~ 2026-09-18 · 19 related posts

• [source] Anthropic redesigns Claude Projects into an orchestrator running parallel cloud threads — claudeai · 2026-09-18
• Claude Projects redesign: brief Claude like a chief of staff with parallel tasks — claudeai · 2026-09-18
• Claude Projects gain shared memory across threads plus a project library — claudeai · 2026-09-18
• Claude project threads run in the cloud; local environment support coming soon — claudeai · 2026-09-18
• [source] Anthropic's Projects redesign: Claude delegates parallel threads, profiles endpoints, opens PRs — claudeai · 2026-09-18
• Anthropic rolls out Claude Code Projects: parallel cloud sessions that split and resume work — ClaudeDevs · 2026-09-18
• Anthropic rolls out Claude Code Projects: one conversation coordinating parallel cloud sessions — ClaudeDevs · 2026-09-18
• [source] Claude Code docs detail Projects: threads run in the cloud and keep going when you're offline — ClaudeDevs · 2026-09-18
• Claude Projects now runs parallel agents from Claude Code that keep working after you log off — testingcatalog · 2026-09-18
• Anthropic Engineer on Claude Code Projects: Built with Projects Itself — maxleiter · 2026-09-18
• Anthropic Launches Claude Code Projects: One Agent Orchestrating Parallel Threads — trq212 · 2026-09-18
• Instagram co-founder Mike Krieger says new Claude Desktop Projects is now his daily driver — mikeyk · 2026-09-18
• Claude Code launches Projects: describe a goal and agents run parallel cloud threads — daniel_mac8 · 2026-09-18
• Claude Code launches Projects; essay argues the agent harness is dissolving into weights — daniel_mac8 · 2026-09-18

5 near-duplicate retellings: lydiahallie · trq212 · bcherny · _catwu · kimmonismus