Six weeks, $1,450, live sessions plus recordings. The brain framework, MCP in depth, plugins and skills, and a capstone that runs on your own tools.
AI Operating Systems is a six-week cohort that moves a daily AI user to a builder. Participants build a brain of context and instructions, connect real tools through MCP, package repeatable work as skills and plugins, and ship one working operating system as a capstone. The programme costs $1,450, one time.
| Layer | What it is | What you build in the cohort |
|---|---|---|
| Brain | The OS itself: context, instructions, memory. Your project brain, rules, voice, and definitions. This is what makes Claude operate as your operator rather than a generic assistant | Weeks 1 and 2 |
| Nervous system | How intent becomes action: the skills, plugins, and templates that do the work on command | Weeks 3 and 4 |
| Legs and hands | The tools, wired in through MCP so Claude executes rather than advises | Week 5 |
| Week | What you build |
|---|---|
| 1 | The brain: context files, operating rules, voice, and the definitions your work depends on |
| 2 | Instructions and memory: global instructions, project memory, and the difference between the two |
| 3 | Skills: packaging a repeatable piece of work so it runs on command instead of being retyped |
| 4 | Plugins: assembling skills into something a colleague can install and run without you |
| 5 | MCP in depth: connecting your live tools so the OS reads and writes where the work actually happens |
| 6 | Capstone: ship one working OS, score it against the rubric, present it to the cohort |
This is the same method behind every operating system we sell. The operating systems SMOrchestra builds for clients at $7,500 per process are built from the six components taught here: a proven framework encoded as skills, the right model routed per task, context files that hold the business, results wired to the live stack, a learning loop, and a scoring gate on every output. A client buying an OS build gets it delivered and transferred in four weeks per process. A participant in T2 builds one themselves in six weeks. The curriculum and the client method are the same thing.
What the capstone has to prove. Your capstone is graded on whether it runs, not on whether it demos. It has to have a named trigger, a named output, a named owner, at least one live tool connection through MCP, and a scoring rubric that gates its output before it ships. That is the same unit definition SMOrchestra puts in client contracts: one process equals one trigger, one output, one owner.
You are an operator, an analyst, a marketer, or technical-adjacent staff. You get good output when you are paying attention and average output when you are not. You have a folder of prompts that other people cannot run. You have been asked to "roll this out to the team" and you know a prompt library will not survive contact with the team.
Who signs, and for how many seats. An individual buys T2 for themselves. A COO buys two or three seats for a small internal team that is about to own a function. Above roughly five seats the arithmetic changes and TC at $12,000 for up to 20 seats is cheaper per head, with the content tailored to the company's own stack.
Skip this if you have never used Claude for real work, in which case T1 is free and is the correct starting point. Or you want the operating system built for you rather than by you, in which case an OS build is $7,500 per process with a two-process minimum. Or you want a revenue engine specifically, in which case T3 at $6,500 builds one on your live pipeline instead of a generic capstone.
| Price | $1,450, one time, no subscription |
| Length | 6 weeks |
| Format | Weekly live session plus recording, cohort community, feedback on the capstone |
| You keep | The OS you built, the skills, the context files, the MCP wiring, and the rubric |
What is not included. 1:1 consulting. Building your OS for you. Tool subscriptions, which sit on your own accounts, because accounts on your side is what makes ownership real.
| What you could do instead | Cost | What you own after |
|---|---|---|
| Buy the operating system built and transferred | $7,500 per process, minimum 2 | The OS, running, with the owner trained |
| Build it yourself in this cohort | $1,450 | The OS, plus the ability to build the next one |
Six weeks, one working operating system, wired to your own tools and yours to keep.
Get the free Claude course, $0, recorded, if you have not used Claude for real work yet.
| Question | Answer |
|---|---|
| Do I need to write code? | No. You need to be able to follow a configuration and read an error message. MCP connections are set up in the cohort, step by step, on your own tools. |
| What tools do I need to have? | A Claude account and at least one business tool you want connected. The connection targets taught in the cohort include CRMs, project tools, and file stores. |
| What if I miss a live session? | Every session is recorded and the community stays open across the six weeks. The capstone deadline is the only fixed date. |
| Is this the same as the revenue engine cohort? | No. T2 builds any operating system on any function. T3 builds one specific thing: a revenue engine on your live pipeline, over 12 weeks, at $6,500, with the GTM diagnostic run on your business first. |
| Will my capstone be a real system or a demo? | Real, and the grading enforces it: named trigger, named output, named owner, one live tool connection, one scoring gate. |
| Can my company buy several seats? | Yes. Above about five seats, TC at $12,000 for up to 20 seats over 6 weeks is cheaper per head and is tailored to your stack, with extra seats at $400. |
| Does this fee credit against a build engagement? | No. Only the two audits and the two diagnostics credit: A1, A2, D1, and D2 credit 100% against any engagement started within 60 days. |
Up: the training ladder. Across: the built-for-you version at $7,500 per process. Down: Start free.