How to Set Up Claude Code So It Never Forgets Your Project
Why AI assistants forget everything between sessions — and the one-time Claude Code setup (memory files + two hooks) that fixes it for good.
July 23, 2026
You know the routine: you open a new chat and start over. “So here’s my project, here’s what we’re working on…” — the same briefing you gave yesterday. And the day before.
Here’s the thing: your AI isn’t dumb. It just has no disk.
RAM vs disk — why AI forgets
Think of Claude like a computer. The context window — its “memory” inside one chat — is RAM: fast, but temporary. Close the session and it’s wiped. What’s missing is a hard drive: permanent storage that survives a restart.
Module 2 of the free Claude Code for Beginners course (Taglish) is about giving it exactly that. Not just an install guide — a setup that learns.
What the setup covers
1. The install order that prevents 90% of stuck installs. Node.js first — before anything else. Then Claude Desktop. Then open the Code tab — and only when it asks, install Git. Most people get stuck simply by doing these out of order.
2. One setup prompt that scaffolds everything.
Paste a single prompt and Claude builds its own system: a self-contained project folder with memory files (lessons-learned, user-preferences, session-handover), a CLAUDE.md operating manual, a secrets folder that never gets committed, and a prepared skills folder. The full prompt is in the video, free.
3. Two hooks — the part most tutorials skip.
Everything written in CLAUDE.md is a reminder — usually followed, sometimes missed. A hook is different: a script the system runs automatically.
- SessionStart reads your memory files every time a session opens — even after a context compact. It never starts from zero.
- Stop counts your conversation turns, and every 10th turn it interrupts Claude and makes it write down anything worth keeping. That’s the difference between a setup that gets older and one that gets smarter.
The proof
At the end of the video I do the only test that matters: I give it a real task, casually mention a preference, then close the entire session. Reopen. No reminders, no re-briefing.
It remembers everything — including the preference I only mentioned in passing.
Watch the full build (Module 2, Taglish): the video above — or search “How to Set Up Claude Code” on YouTube.
Shortcuts if you want them:
- Free Beginner Roadmap checklist — track your progress through the whole course.
- Claude Setup Doctor — the ready-made folder plus a 25-point health check that actually runs your hooks to prove the setup is live. Unzip, tell Claude “let’s get started,” done. Name your price.
Less manual work, more lakwatsa. 🌴