Vigyata.AI
Is this your channel?

I Gave Opus 4.6 an Entire Dev Team - Claude Code Agent Teams

22.6K views· 542 likes· 7:19· Feb 10, 2026

🛍️ Products Mentioned (5)

🚀 Access ALL video resources & get personalized help in my community: https://www.skool.com/agentic-labs 💬 My AI voice-to-text software (Wispr Flow): https://wisprflow.ai/r?LEON114 ☕ Buy me a coffee: https://www.buymeacoffee.com/leonvanzyl 💵 Donate using PayPal: https://www.paypal.com/ncp/payment/EKRQ8QSGV6CWW 📃 Claude Code Agent Teams Documentation: https://code.claude.com/docs/en/agent-teams In this video you will learn how to use Claude Code Agent Teams, a powerful new experimental feature released with Opus 4.6 that lets you create multiple specialized AI agents that work together to build complete applications. Watch as we build a fitness tracker app using a team of five agents including a UX/UI designer, back-end developer, technical architect, database expert and a devil's advocate. You will learn how Agent Teams differ from sub-agents, how to enable Agent Teams in your Claude Code settings.json, and how to use features like delegation mode, human in the loop, and peer-to-peer agent messaging. ⏰ TIMESTAMPS: 00:00 Opus 4.6 Agent Teams building a fitness tracker app 00:42 Watching five specialized agents run in parallel 01:01 Navigating between individual agent team members 01:19 Each agent gets its own Claude Code instance 02:00 Agent Teams vs Sub Agents explained 02:51 How Agent Teams communicate with peer to peer messaging 03:50 When to use Agent Teams vs Sub Agents 04:12 Human in the loop and giving agents instructions 04:39 How to enable Agent Teams in settings.json 05:17 Configuring your prompt for Agent Teams 05:45 Delegation mode and chatting with teammates 06:42 Final app demo and results #claudecode #vibecoding #agenticcoding

About This Video

Opus 4.6 just dropped, and it shipped with a seriously cool experimental feature: Claude Code Agent Teams. In this video I literally give Opus an entire dev team to build a fitness tracker app called Claude Fit from a simple PRD. I spin up five specialized teammates—UX/UI designer, back-end dev, technical architect, database expert, and my favorite: a devil’s advocate that questions everything—then I let them run in parallel while I jump between their individual chats. The big takeaway is how Agent Teams differ from sub-agents. Sub-agents are basically one-way workers: they do a task in their own context and report back to the parent, but they can’t see or coordinate with each other. Agent Teams are more like putting a bunch of specialists in the same room: they share a task list, they can message each other peer-to-peer, and they coordinate automatically—especially useful for complex builds like starting a project from scratch. I also show you how to enable Agent Teams in your Claude Code settings.json, how to prompt for specific roles and models, how human-in-the-loop works (interrupt, redirect, ask questions), and how delegation mode lets the team lead rely purely on teammates. We finish by running the app and the UI honestly looks great—fonts, colors, and even animations feel like a real fitness tracker.

Frequently Asked Questions

🎬 More from Leon van Zyl