agentos.guide › blog

How To Learn Creating AI Agents: Design One Agent First, Then Plug It Into Your Agent OS

By Julian Goldie · 13 October 2026 · agentos.guide

How to learn creating AI agents starts with a one-page brief, because you should design the agent before you touch a tool.

I'm Julian Goldie, and I run a crew of AI agents inside my own Agent Operating System.

The path is short, and it goes like this.

• You learn the nine parts every agent has.
• You fill in a one-page agent brief for one small job.
• You create that agent in a free tool.
• You test it on five real tasks, break it on purpose and fix it.
• You plug it into your Agent OS so it shares memory with your other agents.

This page walks you through one agent from the blank brief to the finished setup.

Quick start: create your first agent today

This is the shortest route from nothing to one working agent.

• First, you pick one small job that you repeat every week.
• Second, you fill in the one-page agent brief further down this page.
• Third, you open a free agent tool, such as StarNet or Hermes Agent.
• Fourth, you create the agent with the brain, personality and approval level from your brief.
• Fifth, you paste in your instructions and switch on one or two tools.
• Sixth, you run five real tasks in chat and score each result.
• Seventh, you change one thing and run the same five tasks again.

You can do the first six steps in one sitting.

The seventh step is the one that turns a toy into something you use.

The anatomy of an agent, and where each part lives in an Agent OS

An AI agent is a model with a job, instructions, tools and something that starts it.

Every agent tool I use asks for the same nine parts, even when the screens look different.

An Agent OS is one dashboard where all your agents live and share one memory.

Mine has three pieces, which are one screen, one shared memory and one task board where agents pick up work.

When you design an agent with these nine parts, it slots straight into that system.

This page is about designing and creating one agent.

My sibling guide on how to learn building AI agents is the project ladder that comes next.

The one-page agent brief

You fill this in before you open any tool.

If the job doesn't fit in one sentence, you aren't ready to create the agent yet.

I teach one agent doing one job, because a small agent is easy to test and easy to fix.

A worked example: one agent, designed first

Here is a brief for a small SEO helper.

SEO is what I told my own StarNet agent I wanted help with, so I've used it as the example.

Every line took a few seconds to write, and the whole brief took a few minutes.

That's the point of designing first, because the tool then only asks you questions you've already answered.

🔥 Want the exact agent creation setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

Create it in StarNet, click by click

StarNet is a free, open-source agent harness that shows your agents in a pixel-art space station.

Its creation screen follows the brief almost line by line.

• You give the agent a name.
• You look at its appearance and change it if you like.
• You pick a personality, and in my video I talked through warm, dry, composed and unhinged.
• You choose the approval level, which is "ask for approval" or "full power".
• You connect a brain by signing in or by pasting an API key.
• You pick the model and the reasoning level, which runs from low through medium and high up to max.
• You create the agent, and a "waking up" animation plays.

I signed in with Grok in one quick step.

I chose Grok 4.6 over Grok 4.7, because I actually prefer Grok 4.6 for that kind of chat.

My agent then said "I'm awake. Let's give the station a direction."

It asked me, "What made you want to set up an AI agent?"

I told it SEO and gave it aiprofitboardroom.com.

After that you open the agent's detail panel.

The panel shows the model, the personality, the instructions, the access permissions and the approval prompts.

You paste your instructions into the instructions field, and you allow or remove access to the web and your local files.

The panel's tabs are Brief, Growth, History and Memories, so you can see what the agent has done and what it has remembered.

I'll keep my honest caveat from the video.

I think StarNet is fun to play with, and I don't think there's a massive amount of practical use cases yet.

There's also a slight delay when you send a message.

It's a very good classroom, though, because every design choice is a button you can see.

I also have StarNet built into my own Agent OS as a tab.

My StarNet AI agent guide covers the rest of the station.

Create the same agent as a Hermes profile

Hermes Agent is an open-source agent from Nous Research that runs on your own machine.

It's the long-runner in my Agent OS, which means it handles the jobs that keep going without me.

You create the same agent with three commands.

hermes profile create scout

hermes -p scout setup

hermes -p scout

The first command makes a new isolated profile.

The second command gives that profile its own model.

The third command opens a chat with it.

A profile is a folder with its own settings, memory and plugins.

I keep one profile per model, so each brain gets a clean agent and I can compare them on the same job.

The persona and instructions go in the profile's SOUL.md file.

The built-in memory is two small files, which are MEMORY.md for the agent's notes and USER.md for your preferences.

My guide to the best Hermes Agent model will help you fill in the brain line.

Design choices and what to pick as a beginner

My rule for permissions is that an agent gets only the access its job needs.

My rule for approvals is that every new agent starts on ask for approval.

I watch what it tries to do for a few sessions, and then I loosen the leash once I trust it.

🔥 Want the exact agent creation setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

How the new agent fits your Agent OS workflow

Creating the agent is the first half of the work.

The second half is deciding how it fits the system around it.

I move every agent through three stages, which are manual, then assisted, then automated.

The three triggers map onto those stages.

Chat is the manual trigger, and in Hermes you start it with hermes -p and the profile name.

A schedule is the automated trigger, and in Hermes you create one with hermes cron create.

A message is the trigger for when you're away from your desk, and hermes gateway setup connects an app such as Telegram.

Here is a real scheduled job from my own guides.

hermes cron create "0 6 * * *" "Read my task list and send me a short to-do recap for today" --name "Daily 6AM To-Do Recap" --deliver telegram

You can read a whole brief out of that one line.

The schedule is the trigger, the prompt is the job and the delivery target is where the result lands.

Hermes also has an emergency stop, which is the hermes pause command.

The shared memory is what makes the agent part of a team.

In my Agent OS the memory is a vault of plain text files, and every agent reads from it.

That means a new agent already knows my business on its first day.

My AI Agent OS guide shows how to build the dashboard, the memory and the task board.

Five agents worth creating first

These five are small, useful and easy to test.

• A notes agent turns a transcript into clean notes with action points.
• A research agent answers one question with a short summary and its sources.
• A keyword scout reads the search queries you paste in and names three to write about.
• A draft-reply agent writes replies to messages you paste in, and you send them yourself.
• A daily recap agent reads your task list and sends you a to-do summary each morning.

Each one has a single outcome that you can check in a minute.

I'd create them in that order, because the first three only need chat and the last one needs a schedule.

How to practise: create, break, fix

You learn creating agents by doing it, and the loop has three moves.

Create

You fill in the brief and create the agent in one sitting.

Break

You then try to make it fail on purpose.

• You give it a vague request and see whether it asks a question or guesses.
• You ask for something outside its job and see whether it refuses.
• You swap the brain for a smaller model and run the same five tasks.
• You lower the reasoning level and compare the answers.
• You remove a permission and watch what it does when a tool is missing.

Fix

You change one instruction, one tool or one setting.

You then run the same five tasks again.

If you change three things at once, you'll never know which change helped.

Every failure points back to a line in the brief, and that's why the brief is worth keeping.

Mistakes I see when people create their first agent

• The job is too big to check.
• The agent starts on full power because asking for approval feels slow.
• Every tool in the catalog is switched on.
• The instructions are a list of keywords, when the model needs full sentences.
• The agent is scheduled before anyone has watched it work in chat.
• The model gets the blame when the brief was the real problem.

Each of these is a design mistake, which is why I teach design before tools.

If the word agent is still new to you, start with how to learn AI agents from scratch.

If you want a structured course, my post on the best AI agent course for beginners compares the options.

My free community, the AI Money Lab, is a good place to share your first brief.

FAQ

What is the best way to learn creating AI agents?

Design one agent on paper first. Fill in a one-page brief covering the job, the brain, the instructions, the tools, the permissions, the memory, the trigger and the approval level. Then create it in a free tool and test it on five real tasks.

Do I need to code to create an AI agent?

No. StarNet creates an agent through a few clicks, and Hermes Agent needs three short commands. The real skill is writing a clear job and clear instructions.

What goes in an agent brief?

An agent brief lists the name, the one job, what finished looks like, the brain, the reasoning level, the personality, the instructions, the hard limits, the tools, the permissions, the memory, the trigger, the approval level and five test tasks.

How does a new agent fit into an Agent OS?

The job becomes a card on the task board, the brain is its own profile, and the memory is the shared vault that every agent reads. The agent moves from manual chat to assisted work and then to a schedule.

Should my first agent run on full power?

No. Start every new agent on ask for approval, watch it for a few sessions and widen its freedom only after it passes your tests.

Which agents should I create first?

Start with a notes agent, a research agent or a keyword scout, because each has one outcome you can check in a minute. Add a draft-reply agent and a scheduled daily recap after that.

Final word on how to learn creating AI agents

Write the brief, create one agent, break it and fix it, and you'll have the real answer to how to learn creating AI agents.

Related guides on agentos.guide

→ How to Build Your Own AI Agent OS

→ StarNet AI Agent

→ Best Hermes Agent Model

→ Hermes Crash Course

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

← all guides on agentos.guide