Full Tutorial: Build an App with Multiple AI Agents and Claude Code | Kieran Klaassen
Move beyond vibe coding to learn how to use Claude Code to build with multiple AI agents. Plus slash commands to save time, how Kieran codes with his voice, and more.
Dear subscribers,
Today, I want to share a new episode with Kieran Klaassen.
Kieran is the founder of Cora, a beautiful AI email assistant. While most of us are still learning vibe coding, Kieran has figured how to use 3 AI agents to build in parallel. I got him to show me exactly how he does this using Claude Code. Watch as he builds an email app in 20 min without typing a single line of code.
Watch now on YouTube, Apple, and Spotify.
Kieran and I talked about:
(00:00) Why agentic coding beats vibe coding
(03:21) "It's like I'm a manager and I have an AI agent team"
(06:29) Why Claude Code is winning over Cursor for power users
(12:15) Live demo: Building a new feature with 3 AI agents in parallel
(22:32) Custom slash commands to eliminate repetitive work
(36:13) AI code reviews using multiple perspectives and bots
(54:55) Inside Cora: Your AI Chief of Staff for email
(1:00:11) Advice for non-engineers intimidated by agentic coding
A more practical focus for Behind the Craft
After 70 episodes, I've found that the best ones focus on practical tutorials that you can follow to save time with AI and build in-demand skills. Upcoming episodes will feature topics that I want to learn myself:
Full Tutorial: Generate Amazing Designs with AI | Meng To
A PM's Guide to AI Evaluations (Step by Step) | Aman Khan
Learn Zapier AI in 30 Min to Automate Your Boring Work | Wade Foster
Subscribe to my YouTube to learn with me — I appreciate your support!
Top 10 takeaways I learned from this episode
Claude Code gives you focus and flexibility beyond traditional IDEs. "It gives me peace because there's just one thing I can do here really, which is type what I want it to do." The terminal interface removes distractions and integrates with everything through MCPs.
Use "think ultra hard" prompts for complex tasks that need quality. Trigger words like "think ultra hard" or "think deeply" increase token usage but improve AI output quality. Reserve them for tech architecture decisions, not simple line changes.