// The Starter · Free

You're the whole marketing team.
Run it like one.

You know what good looks like. What you do not have is anyone to produce it, in every discipline the job covers, at the volume it asks for.

Marketing Team of One is an operating system for your AI agents. It gives them what a good team has: your business context, your standards, and a memory of every decision. Every agent starts the day knowing your business, and you spend yours directing the work.

Everything lives in one folder, and Claude, Codex, Grok Bot, and Hermes Agent all work from it.

Get the free Starter See what is inside

Start free. The Starter installs in under thirty minutes and it is genuinely useful on its own.

Worked example included Works with Claude, Codex, Grok Bot, and Hermes No account beyond your agent
// 01 · The situation

Two ways people end up here.

Path one

You have less to work with than you used to. You are being asked for more.

Maybe the team was cut underneath you. Maybe you took a new role and found a fraction of the resources you expected. Once you are in it, the route stops mattering.

You know exactly what good looks like. You have commissioned it, reviewed it, and sent it back until it was right. That does not go away when the headcount does.

What you are short of is the capacity to produce it. And there is someone watching to see whether you deliver anyway.

Path two

You are the first marketer. There is nobody to hand the work to.

Nothing was taken from you, because there was never anything there. You are the strategy, the campaigns, the writing, the design and the reporting.

You know what good looks like. What you do not have is anyone, or anything, to produce it with you. Every piece of it waits on you personally.

Different routes. Same position. You are the one who knows what good looks like, and there is nobody else to do the work.

Some of you did not choose this, and nothing here will pretend that was an upgrade. Some of you did choose it, and would not go back. From here the job is identical.

There is nobody to escalate to. Depending on how you got here, there may not even be anyone to blame. What this gives you is an execution layer that holds your standard instead of diluting it, and a system that remembers the decisions so you are not making them again every week.

// 02 · The problem

The model is fine. The setup is the problem.

You already use AI. You are also re-explaining your business to it every single time, and the output quality tracks how well you explained it that day. Nothing accumulates.

Working in chat
  • Re-explain the company, the audience, the voice, every session
  • Quality swings depending on how you framed it that day
  • Decisions live in a chat you will never find again
  • No operating cadence, so marketing runs on whatever is loudest
  • Every good output is a one-off you cannot reproduce
Working in the system
  • Your agent reads your business before it answers, every time
  • Same context in, consistent quality out
  • Decisions and lessons accumulate in a file that carries forward
  • A weekly rhythm that sets priorities and closes the loop
  • Skills that produce the same shape of work every time you ask
// 03 · What you install

A folder your agents read before they answer.

Not a prompt pack. Not a course. A working folder with a context layer, four skills, and an operating rhythm, plus a complete example so you can see what finished looks like.

Layer 01

Context

Three files holding what your business is, who buys, and how the brand sounds. Stable knowledge your agent loads on every request.

Layer 02

State

What is actually in flight this week. Priorities, blockers, what you are waiting on. Updated in the weekly close.

Layer 03

Memory

Decisions, what worked, what did not, and the corrections you are tired of repeating. Append only. This is what compounds.

The four skills
Say thisWhat happens
run onboarding Interviews you for about ten minutes and fills in your context from your answers.
campaign brief for X Turns a one-line ask into a scoped brief: goal, audience, message, channels, sequence, success measure.
content brief for X Turns a topic into a brief a writer or an agent can execute without coming back with questions.
check this QAs any draft against your own positioning, your audience's language, and your voice. A quality gate that knows your standard, not a general one.
The rhythm

Start Monday clear. End Friday with proof.

Fifteen minutes on Monday: rank the week against your quarter goals and decide what to turn down. Fifteen minutes on Friday: record what shipped, what worked, and what to stop. When someone asks what marketing did this month, the answer is already written. Every Friday makes the next Monday sharper.

The example

See it finished before you start.

The hardest part of any new system is picturing it working. So the Starter ships with a company that already runs on it: a fictional B2B business with its context filled in, nine months of decisions, four honest failures, and a campaign brief the system wrote from one line. Read it first. You will know exactly what yours is becoming.

// 04 · The install

Under thirty minutes, start to working.

1. Enter your email and download the folder.

2. Unzip it somewhere you will find it again.

3. Open the folder in your agent: Claude Cowork, Claude Code, Codex, Grok Bot, or Hermes Agent.

4. Say run onboarding and answer the questions.

5. Ask for a campaign brief and watch it use what it just learned.

You need your agent on a paid plan, or your own model key for Hermes Agent. In Claude Cowork there is no API key, no terminal, and no developer setup. Full walkthrough: How-to.

What lands in the folder 
context/     business · audience · positioning
state/       now
memory/      decisions
workflows/   weekly rhythm
example/     a finished company
CLAUDE.md    how Claude runs this workspace
README       setup + troubleshooting
// Free · No card

Get the Starter

Enter your email and the download is on the next page. You will also get The Dispatch, a weekly note on what changed in the agents you use and what shipped to the OS. Unsubscribe from any email.

Your context files live on your machine, and the Starter itself sends nothing anywhere. What your agent reads goes to its provider under their terms.

// 05 · Scope

What the Starter does not do.

Named plainly, because a free thing that oversells itself costs more trust than it earns.

The Starter is a working foundation, not a full function. It is deliberately small enough that you actually finish setting it up, which is more than most systems manage.

// 06 · Still deciding

Find out what you are leaving on the table.

Maybe chat is working well enough that none of this feels urgent. That is worth testing rather than assuming.

The self-assessment is twenty plain-English questions about how you actually work with Claude today. Nothing in it assumes you are technical. It measures how you operate, maps what you have not touched yet, and shows the distance between the two: where your setup stands, your biggest gap, and how much is still sitting there to unlock. It ends with three moves sized by the time you actually have, one for today, and text you can paste straight into Claude. No email gate on the results, and your answers never leave the page.

If it tells you your setup is already ahead, believe it and keep what you have. If not, you will know exactly where to start.

Take the assessment
// Coming next

The complete Team of One OS

The full system: a broad marketing skill library, work mapping that sorts what stays human-led from what becomes AI-assisted, a configurable governance layer, and twelve months of updates so it keeps working as the agents change.

Founding pricing for the first cohort. Get the Starter and you will hear before it opens, no separate signup needed.

No date promised. No card. You will not be sold to in the meantime.