Folder structure for AI agents — the rule that saves your sanity

Your folder structure is why your agents are confused.

The walkthrough

1. Your folder structure is why your agents are confused.
2. Mixed projects. Mixed context. Agent picks the wrong one every time.
3. The fix is dumb-simple.
4. One folder per project. No exceptions.
5. Each one gets a CLAUDE.md at the root. That's the rule.
6. Three sections. What it is. The stack. The status.
7. Plus your rules. How you want it to behave in THIS folder.
8. Every agent reads this file before it touches anything else.
9. Without it: 200 lines that don't match your intent. With it: clean.
10. This is mine. Every project has its own folder. Every one has CLAUDE.md.
11. Even the parent folder has one. Inherits down.
12. Cursor and Claude Code both pick this up automatically.
13. This is where you write 'never use jQuery' or 'always SQLite.' One place.
14. Sensitive stuff in its own folder. Agent's CLAUDE.md says 'never read this'.
15. Update the file once. Every future session starts with that context.
16. Scales to as many projects as you want. Same rule every time.
17. Knowledge stays in one place. Tools all read from it.
18. Your folder is the moat. CLAUDE.md is the front door.
19. localmodeldojo.com — Founder pricing $24/mo for life · code LAUNCH30 · 30 days only

— Jake Morris · Oklahoma veteran · localmodeldojo.com