Simon Scrapes

Anthropic Just Built It’s Own Agentic OS

Sep 18, 2026 8 min
anthropicclaudeagentic workflowsai agents
Watch on YouTube Follow Simon Scrapes on Rundown — free

Summary

AI summaries can be incomplete or wrong. Verify anything important against the original video.

This video breaks down Anthropic's recent update to its 'Projects' feature, effectively positioning it as a foundational layer for an agentic operating system.

The video analyzes how Anthropic has enhanced its 'Projects' feature in Claude, evolving it into a more sophisticated, long-running agentic tool. Key improvements include a new retrieval-augmented generation (RAG) pipeline for smarter context management, refined cross-thread collaboration, and enhanced permission controls for teams. These changes move Anthropic toward an 'agentic operating system' (OS) that can coordinate multiple sub-agents under a single, ongoing conversation. The speaker discusses why these architectural shifts are significant for managing complex workflows, while noting that the system is still an early version of this vision.

Verdict

Claude Projects
ai software

Anthropic has implemented a crucial architectural shift, building a scalable foundation for agentic workflows rather than just adding features.

Buy

Pros

  • ✓ RAG enables efficient context management beyond the basic context window limits. 7:15
  • ✓ Unified project overview pane improves coordination across multiple sub-threads. 5:25
  • ✓ New sharing and permission controls allow for true team-based collaboration. 5:29

Cons

  • ✗ Currently limited to Pro, Team, and Enterprise plans. 3:47
  • ✗ Features are currently in a rolling release, not yet available to all users. 3:49
  • ✗ The main agent conversation still lacks direct access to external MCP tools, requiring thread-level delegation. 7:21

Best for

  • power users
  • software engineering teams
  • project managers

Key Points

  • 1:53 Projects now function as long-running, orchestratable conversations that delegate tasks to multiple sub-agents.
  • 5:46 New team-level permissions and sharing options allow multiple users to collaborate on the same project and context.
  • 7:07 Anthropic added RAG to projects, allowing them to effectively manage large amounts of context by retrieving only necessary information.
  • 7:32 The speaker argues this update represents the foundational framework, rather than the final version, of an agentic operating system.

Worth watching if: You are a power user of Claude or interested in the architecture of agentic workflows and how AI platforms are evolving to support team-based collaboration.

Get every Simon Scrapes video extracted like this

One daily email with structured extracts of every channel you follow. Free tier covers 15 videos a month.

Sign in with Google

No credit card. Free tier forever.

Watch on YouTube