🚀 Start Here

Setup & Context OS

Everything you need to get your first AIOS workspace running — the template structure, the four context files that give Claude its intelligence, the core commands, and the session workflow you'll use every single day.

📽 Workshop recording
~25 min read
Foundation — do this first
Includes template walkthrough
Section 01

What Is an AIOS Workspace?

A workspace is just a folder on your computer. That's it. Inside that folder live the context files, commands, scripts, and outputs that turn Claude Code from a blank AI into something that knows your business, your role, your strategy, and your data.

You open the folder in VS Code (or any IDE), open a terminal, and start a Claude Code session. Claude reads the workspace, loads your context, and is immediately oriented — who you are, what you're trying to achieve, where everything lives. Every session starts fresh but never blank.

💡
Context is everything. Without a properly filled context OS, every Claude session starts from zero — you're re-explaining yourself constantly. With it, Claude walks in already knowing your business, your strategy, your current numbers, and your role. That's the difference between a general-purpose chatbot and a business operating system.
Alternative
Claude Desktop App
  • + Easier for non-technical users
  • + Built-in Claude Code integration
  • ~ UI limits on context processing
  • ~ Less control over automation
  • ~ Ceiling on what you can build

Cursor works the same way — same folder, same commands, same approach. The workspace is platform-agnostic. You can move it anywhere.

Section 02

The Workspace Template — File by File

The starter template gives you the scaffold. Download the ZIP, unzip it, rename the folder to something that represents your business (e.g., advisoraipartners-workspace), and open it in VS Code. Here's what's inside and what each piece does.

your-workspace/
├── CLAUDE.md        ← Loaded every session. The source of truth.
├── .claude/
├── commands/      ← Slash commands (/prime, /create-plan, /implement)
├── prime.md
├── create-plan.md
└── implement.md
└── skills/        ← Heavier plugins with folders, references, scripts
├── mcp-integration/
└── skill-creator/
├── context/        ← The Context OS. Fill these in first.
├── business-info.md
├── personal-info.md
├── strategy.md
└── current-data.md
├── plans/         ← Plans created by /create-plan
├── outputs/       ← Deliverables, analyses, work products
├── reference/     ← Templates, patterns, reusable materials
└── scripts/       ← Automation scripts and cron jobs
📁
Hidden files on Mac: The .claude/ folder starts with a dot, so it's hidden by default in Finder. Press Cmd + Shift + Period to toggle hidden files on and off. You'll need to do this once after downloading the template to confirm everything is there.
Section 03

CLAUDE.md — The Foundation Document

CLAUDE.md is automatically loaded at the start of every Claude Code session. It's the first thing Claude reads — the document that tells it what this workspace is, where everything lives, and how to operate within it. Think of it as the orientation brief Claude gets every time it walks in the door.

Always-loaded context: Every session, Claude reads CLAUDE.md first. It knows the workspace structure, what commands are available, and the rules for maintaining the workspace — before you've typed a single word.
Living document: Whenever Claude makes a significant change to the workspace — adding a new command, creating a new output type, changing the structure — it must update CLAUDE.md to reflect that. The instruction is baked into the template itself.
What it contains: The workspace structure (as a directory map), descriptions of each key directory, the available slash commands, the session workflow, and the rule that CLAUDE.md must always match reality. Keep it current — stale CLAUDE.md = confused sessions.
Don't bloat it: CLAUDE.md is loaded every session, so size matters. Keep it accurate and minimal. Deep detail belongs in the context files or command files, not here. This is the map, not the territory.
⚙️
The self-maintaining workspace: Because CLAUDE.md instructs Claude to update it whenever the workspace changes, the workspace becomes self-documenting. Add a new command and Claude will add it to CLAUDE.md without being told. This is what makes the workspace stay coherent across hundreds of sessions.
Section 04

Context OS — The Four Files

The context/ folder is the heart of the AIOS setup. These four files are what Claude reads during /prime to fully orient itself. Filling them in is the single most important thing you do when setting up a workspace — for yourself or a client. Everything else builds on top of this foundation.

Each file has a specific job. Together they answer the four questions every AI assistant needs answered to be genuinely useful: What is this business? Who am I helping? What are we trying to achieve? Where do things stand right now?

🏢
business-info.md
Organizational context — what and who
Sets the stage for everything else. Describes the organization — what it does, who it serves, its products or services, stage, team size, and any important constraints. Claude reads this to understand what kind of work it's supporting before it reads anything else about your role or goals.
Organization Overview What does the business do? Who does it serve? What's its mission or focus?
Products / Services Primary offerings or areas. Helps Claude understand what you're building or delivering.
Key Context Stage, team size, market position, recent changes, or constraints Claude should know.
👤
personal-info.md
Your role — who Claude is helping
Defines your place within the business. Claude needs to know who it's talking to — your title, what you're accountable for, what decisions you make, what your daily work looks like. This is what lets Claude tailor every response and output to your specific situation rather than giving generic advice.
Your Role Title, function, primary responsibilities. What are you accountable for?
Key Responsibilities Core areas you manage. Helps Claude understand what tasks are relevant.
How This Workspace Helps What recurring tasks, analyses, or decisions do you want Claude's help with?
🎯
strategy.md
Current priorities — where you're going
Captures what you're actually working toward right now. This is where Claude gets aligned with your direction — not just what the business does, but what it's optimizing for this quarter or year. Without this, Claude is helpful but not strategic. With it, every suggestion and output is pointed in the right direction.
Current Focus Period What timeframe does this strategy cover? Q1? Full year? A specific initiative?
Strategic Priorities The 2–5 things you're actually focused on achieving. Be specific.
What Success Looks Like Outcomes and milestones that define "done" for this period.
Key Decisions or Open Questions Strategic tensions or trade-offs Claude should be aware of.
📊
current-data.md
Live numbers — the chef's kiss on top
The numbers behind the narrative. Revenue, users, conversion rates, pipeline status, team capacity, recent wins and losses — anything Claude needs to do real analysis rather than generic advice. Starts as a manually updated snapshot; eventually becomes a live file refreshed by scripts pulling from your actual data sources.
Key Metrics A table: Metric | Current Value | Target | Notes. Fill in what matters most.
Current State Qualitative: status of key projects, recent wins/losses, blockers, capacity.
Data Sources Where this data comes from — useful for building automation scripts later.
🔗
How they connect: business-info sets the organizational stage → personal-info defines your place in it → strategy.md says what you're optimizing toward → current-data gives Claude the real numbers. Each file references the others so Claude sees the complete picture, not four disconnected documents. Together they become the intelligence layer that makes everything else possible.
📈
Automate current-data over time: Once you're comfortable, add a script to scripts/ that pulls from your real data sources — YouTube Analytics, Google Analytics, Stripe, your CRM, your P&L — and refreshes current-data.md automatically. That's when the daily brief and business intelligence features really come alive.
Section 05

Core Commands — The Three Workhorses

Commands live in .claude/commands/ as markdown files. They're invoked with a slash prefix in your Claude Code session. The template ships with three that you'll use in every session — think of them as the engine, not the content. They work across any task, with any context.

/prime
Session initialization
Run first, every time
Purpose: Load all context and orient Claude for the session. Run this at the start of every single session — it's the ritual that makes the workspace system work. Claude will scan the workspace, read your context files, and confirm back: who you are, what the workspace is for, what commands are available, your current strategy, and a confirmation it's ready to help.
→ Run ls, read CLAUDE.md and context/ → summarize understanding → confirm ready
/create-plan
Plan before you build
Use before any substantial change
Purpose: Before making significant changes — adding a command, building a new workflow, restructuring the workspace — use this to create a thorough plan. The plan lives in plans/ with a dated filename so you have a history. Claude will think through the problem, document the approach, and list step-by-step tasks before touching anything.
/create-plan add a competitor analysis command for my podcast
/implement
Execute a plan
Use after /create-plan
Purpose: Takes a plan created by /create-plan and executes it step by step. Reads the plan, works through each task in order, validates the work, and updates the plan status. These two commands are designed to be used together — plan first, implement second. This separation prevents half-baked changes and keeps a record of intent.
/implement plans/2026-04-25-competitor-analysis-command.md

Beyond these three, you'll build your own commands over time — /data-audit, /daily-brief, /analyze-competitor, whatever your recurring workflows demand. Each one is just a markdown file in .claude/commands/ that tells Claude what to do step by step. That's the entire mechanism.

Section 06

Commands vs. Skills — What's the Difference?

Claude Code technically calls both "skills," but in practice they're two different things with different weights. Getting clear on the distinction helps you know what to reach for when you're building new functionality.

Commands vs. Skills — a practical distinction
Commands Lightweight
  • A single markdown file in .claude/commands/
  • Invocable with a slash command (/prime, /daily-brief)
  • Works as the glue that stitches together context, data, and functionality in a sequence of steps
  • Can reference scripts and other files within the workspace
  • Good for: workflows you want to invoke manually on-demand or run on a schedule
Skills Heavyweight
  • A full folder in .claude/skills/ with a SKILL.md entry point
  • Contains subfolders: examples/, references/, scripts/
  • Teaches Claude a new capability — design patterns, API access, tool usage — not just a workflow
  • Can be user-invocable or non-invocable (background capability)
  • Good for: giving Claude access to external systems, APIs, or specialized domain knowledge
🧩
In practice: A command says "do these steps in this order." A skill says "here's how to use this tool or think about this domain." Your /daily-brief command might use the MCP integration skill to pull data from external sources — the command orchestrates, the skill provides the capability. They're designed to work together.
Section 07

Getting Set Up — Step by Step

Here's the exact sequence for getting your first workspace running. Most of you will fly through steps 1–4. The time investment is in step 5 — filling in your context. That's where the real value gets created.

1
Install VS Code (or Cursor)
Download from code.visualstudio.com. If you already have it, skip ahead. Cursor works identically — same folder structure, same commands, same approach.
2
Install Claude Code CLI
Open a terminal in VS Code (Terminal → New Terminal) and run the install command for your OS.
npm install -g @anthropic-ai/claude-code
You'll need Node.js installed. If you don't have it: brew install node (Mac) or download from nodejs.org.
3
Download and unzip the template
Get the starter template ZIP from the course materials. Unzip it and rename the folder to something that represents your business.
e.g. → advisoraipartners-workspace/
Mac: Press Cmd+Shift+Period to reveal hidden files after unzipping — the .claude/ folder will be invisible otherwise.
4
Open the folder in VS Code and start Claude Code
File → Open Folder, select your workspace. Then open a terminal and start a Claude Code session.
claude
Claude Code will authenticate with your Anthropic account. Recommended plan: Max ($200/mo) for heavy use, or start with Pro ($20/mo) and upgrade when you hit limits.
5
Fill in your Context OS files
Open each file in context/ and fill in your information. Start with business-info.md, then personal-info.md, then strategy.md. current-data.md can wait until you have real numbers to put in.
context/business-info.md → context/personal-info.md → context/strategy.md
Don't worry about perfection. A rough draft is infinitely better than nothing. Claude will get dramatically smarter with each sentence you add.
6
Run /prime and confirm it works
In your Claude Code session, type the prime command. Claude will read your workspace and summarize back what it understands about you, your business, and your goals. If it sounds right, you're ready to go.
/prime
Claude will list: who you are, what the workspace is for, available commands, your current strategy, and a readiness confirmation.
Section 08

Your Daily Session Workflow

Once the workspace is set up, every day looks the same. The ritual is the same whether you're doing a 15-minute task or a 4-hour build session. Consistency here is what makes the workspace feel like a real operating system rather than a chatbot.

1
Start
Open workspace in VS Code, open terminal
claude
2
Prime
Load context, orient Claude for the session
/prime
3
Work
Use commands or direct Claude with tasks in natural language
4
Plan
For anything substantial, plan it before building
/create-plan
5
Execute
Implement the plan step by step
/implement
6
Maintain
Claude updates CLAUDE.md and context/ as the workspace evolves
One workspace, keep building layers: You don't need a new workspace for every project. Build layers on top of the same workspace. Add new commands. Add new data sources. Add context as your business evolves. The workspace grows with you — that's what makes it an operating system rather than a one-off tool. Separate workspaces only when you need separate contexts entirely (e.g., a completely different business, or a team that shouldn't see your full workspace).
Tips & Gotchas

Things Worth Knowing

Show hidden files first. On Mac, Cmd + Shift + Period toggles hidden files in Finder. You need this to see the .claude/ folder after downloading the template. Easy to miss; causes a lot of confusion when people think the template is empty.
/prime every session, no exceptions. Claude Code sessions don't carry state between them. If you skip /prime, Claude is working without your context — it'll give generic responses instead of business-aware ones. Make it automatic.
Fill in context before you do anything else. The instinct is to start building commands and workflows immediately. Don't. Empty context files mean every output is generic. Even rough, incomplete context files make a massive difference. Do those first.
Max plan is worth it. The $200/month Max plan removes the usage friction that derails deep work sessions. If you're using this for your business, the first good session will pay for the month. The $20/mo Pro plan works — upgrade when you start hitting limits regularly.
Commands vs. the VS Code sidebar. Some people use the Claude Code panel built into VS Code rather than the terminal. The terminal approach (recommended here) is faster and can process more context. The UI is limited by its interface. Use the terminal.
current-data.md starts manual, becomes automated. Fill it in manually at first — even rough numbers are better than nothing. As you get comfortable, add scripts to scripts/ that pull from your actual data sources and keep this file fresh automatically.
CLAUDE.md must always match reality. If you add a new command but don't update CLAUDE.md, future sessions won't know it exists. Claude is instructed to update this file after any structural change — but always worth checking if something feels like it's being forgotten.
Multiple terminals = multiple sessions. You can run several Claude Code sessions in parallel by opening multiple terminals in VS Code. Each session is isolated. Useful for running one session for focused work while another handles a background task.
What Comes Next

The Learning Path from Here

Context OS is the foundation. Data OS builds on top of it. Once you have both running, you start adding commands, then automating them. Here's the week-by-week sequence these workshops were originally designed around.

Day 1 (this module): Setup + Context OS — get the workspace running and fill in your four context files.
Day 2 — Data OS: Integrate your real data sources. P&L, YouTube Analytics, CRM, Stripe, Bitly. Build the local database layer that makes current-data.md live and automatic.
Day 3 — Custom Workflows: Build your first custom commands. Create a /data-audit, a /daily-brief, or any recurring workflow you currently do manually. Use /create-plan + /implement to build them safely.
Day 4 — Recurring Tasks & Cron: Wire your best workflows to triggers and schedules. Work done while you sleep — the daily brief, the data refresh, the status check. This is the delegation layer.
Day 5 and beyond — AIOS as a Service: Take what you've built for yourself and start offering it to clients. Refer to the Builder Mindset module for the delivery process and pricing framework.
Continue Learning

Related Modules