The AI OS Starter

What an AI Operating System actually is

Most people use AI like a search box. You ask a question, you get an answer, you copy-paste it somewhere, you move on. Nothing is saved. Nothing compounds. Tomorrow you start from zero again.

An AI Operating System is the opposite. It is a persistent, configured environment where the AI already knows your business, your tools, your rules, and your goals, and can take real action across all of them. You stop prompting AI and start operating it.

It comes down to three layers:

• The kernel (CLAUDE.md) - the file the AI reads at the start of every session. Your security rules, your tools, your preferences, and your lessons learned all live here. This one file is what turns a generic assistant into a configured employee.
• The skills (.claude/skills/) - repeatable workflows you invoke by name. Each one is a job your AI does the same way every single time.
• The connections - the wiring to your calendar, CRM, content tools, and data, through API keys and MCP servers.

The AI OS Starter gives you all three, pre-built, so you can have a working system today instead of in three months.

The 5 core skills

These five cover the most repeated, highest-leverage workflows in almost any business. Between them they remove hours of manual work a day. Each one lives in its own folder and is built to be edited - open it, read it, make it yours.

1. /daily-brief

Pulls your calendar, open tasks, and key numbers into one tight morning brief, then ends with the single highest-leverage action for the day. Replaces the 20 minutes you spend every morning piecing together what is actually going on.

2. /content-machine

Researches your niche first, then writes a batch of short-form posts or scripts in your voice. Demo-first hooks, one idea per piece, your tone rules enforced. A week of content from one session.

3. /lead-engine

Captures new leads, qualifies them against your criteria, and drafts personalised follow-ups in your voice. Runs a follow-up sweep across your pipeline so no lead goes cold. The money is in the follow-up, and this is the part most people skip.

4. /meeting-prep

Before any call, it builds a one-page brief from your calendar, past notes and transcripts, your CRM, and a quick web lookup. You walk into every meeting already knowing who you are talking to and what they want.

5. /deep-research

Fans out parallel research agents across a hard question, then synthesises one clear, decisive answer with sources. Fast model for the legwork, strong model for the judgement. Market research and big decisions in minutes, not afternoons.

Install it in about 5 minutes

The whole thing lives in one GitHub repo. Here is the entire setup.

Step 1 - Get Claude Code

If you do not have it yet:

New to Claude Code entirely? Read Install Claude Code first, then come back.

Step 2 - Clone the AI OS Starter

Step 3 - Add your keys

Step 4 - Make the kernel yours

Open CLAUDE.md and replace the bracketed placeholders with your real business, tools, and rules. This is the single most important step. The more honest and specific this file is, the better your whole AI OS works.

Step 5 - Run it

Then type:

That is it. You now have a working AI Operating System.

How to make it grow

The starter is a foundation, not a finish line. Three ways it compounds:

• Add skills. Copy any folder in .claude/skills/, rename it, rewrite the SKILL.md. The folder name becomes the command. Learn the pattern in Build Skills.
• Connect tools. Add an MCP server for your calendar, CRM, or content stack, then list it in CLAUDE.md so the AI knows it exists. See Connect MCP Servers.
• Teach it. Every time the AI gets something wrong, add a line to the Known Failures section of CLAUDE.md. It never repeats the mistake. That section is your real moat over time.