Building a multi-agent MCP product with a dev team. We're debating architecture at the routing/state layer. Our current design uses layered index.md files — ICM-style. The top-level index routes query type, agent-specific indexes define stage contracts, Layer 3 holds the stable reference files, Layer 4 is the per-run working content. Clean separation, low token overhead, human-readable at every step. The question on the table: LangGraph vs. staying with the index.md architecture. The argument for LangGraph is state management — persisting data between agents mid-conversation without passing full JSON payloads downstream. The argument against is that we may not have that problem yet, and adding a framework dependency before the problem is real creates overhead without payoff. Jake's paper draws a hard line on where ICM does and doesn't work. Complex automated branching mid-pipeline is one of the failure modes flaged. Is that the trigger? Meaning: if agents need to share state mid-run rather than just receive the previous agent's output, is that the point where a framework like LangGraph earns its weight over a file-based architecture? Or is there a hybrid approach — ICM for context scoping, LangGraph only for the state graph layer — worth considering before we commit either direction?

Hey everyone, If you’ve been quietly struggling with learning how to build or what to build, you’re not alone. I’ve seen the posts, and I know how frustrating it can feel when things don’t click. I wanted to try to do something about it. I built a custom AI Tutor that has the complete knowledge and teaching style of @Jake Van Clief and @Matthew Creamer’s Foundations courses. Think of Jake as the professor. This agent is your dedicated Teacher’s Assistant who sits with you one-on-one. Here’s why it’s different: • It walks you through every concept step by step, exactly as Jake outlines them • You’re not just chatting or watching videos. You’re actually building files and folders together in Claude • It checks your work in real time and won’t let you move forward until you truly understand • It speaks in their voice and follows their exact methodology This is the hands-on practice and patient guidance a lot of us have been missing. How to try it: 1. Grab the repo 2. Follow the README 3. Open the folder in Claude and let the tutor take over I’m very proud of how it turned out and I want this to help as many people as possible. We just crossed 30,000 members! Let’s make sure every single one of us has the support they need to actually master this stuff. If you try it, please drop a comment and let me know: • Did it help you understand a concept that was confusing before? • What worked well? • What could be better? I want your honest feedback so we can keep improving it together. I’d love it if this could turn into a proper community teaching tool we all build and evolve. Link in comments 👇