A-TIER guide
Agent workspace file tree — persistent memory and context folder layout
100k ft — What is this?
Agent workspace file tree — persistent memory and context folder layout
50k ft — Why care?
A folder structure for Claude-based agents: a brain folder with CLAUDE.md and model routing, five business context files, a memory folder with session logs and an index, per-project context files, NOW.md short-term memory, skills/workflows/scripts folders, Obsidian + GitHub versioning, Notion tasks, and secrets in a gitignored .env.
10k ft — Key takeaways
• Give agents a stable file tree so memory survives across sessions
• Keep a per-project context file plus a short-term NOW file
• Put secrets in .env and gitignore it before the first push
5k ft — Tools & techniques
ChatGPT, Claude, Claude Code, GitHub, Notion, Obsidian
AI agents, agent memory, automation, context files, model routing
1k ft — Deep-dive / try it
Comment “FILES” and I’ll send you the full file tree + the prompts that build it 👇
Most people kick off an agent by asking it to do something.
The ones that actually work start with the folder structure, because an agent with no file structure forgets everything, and an agent that forgets starts every session from zero.
This is the exact structure I set up before any agent gets built: where the knowledge lives, what it remembers, why every project needs its own home, and what goes in .gitignore before you ever push.
#ai #claude #chatgpt #agent #automation
Transcript excerpt
Have one main folder that holds all your projects, a brain folder inside that with a Cloud MD at the top with all your rules and preferences. Set up model routing to save on tokens and have five main context files on your business, offer, customers, voice and goals. Have a memory folder to keep track of all your sessions, an index file to keep track of where everything is, a tools MD file with all your connected apps, project folders for each thing you're working on, a context file in each so it knows where everything stands, all of that connected to Obsidian then linked to GitHub for version control. Then set up a now MD file for short-term memory, connected to Notion to organize open tasks for each project across sessions, and a rules file to organize everything so projects don't bleed together. Have all your skills in one folder, all your workflows in another, then a folder with scripts for repeat tasks to save on token usage. Have your passwords and keys in one .env file and get ignore so everything stays secure. And if you comment files, I'll send you the whole breakdown, plus a guide with every prompt so you can build this yourself, and follow so you can keep up with everything AI.