agentos.guide › blog
How To Learn AI Agents: The 8-Step Path That Ends With Your Own Agent OS
How to learn AI agents is an eight-step path, and the fastest way through it is to treat each step as one layer of your own Agent OS.
You understand what an agent is, use one ready-made agent, give it tools, add memory, schedule it, teach it skills, run several agents and then build the dashboard.
I'm Julian Goldie, and I run my business on a crew of agents inside an Agent Operating System that I built this way.
This page gives you the steps in order, a quick-start you can run tonight, a four-week plan and a free resource for every step.
How to learn AI agents: the eight steps, and what each one builds
This is the complete path in order, and each step leaves a working piece of an Agent OS behind.
| Step | What you learn | The Agent OS layer it becomes |
|---|---|---|
| 1. Understand what an agent is | You learn the difference between a chatbot and an agent. | It is the idea behind the whole system. |
| 2. Use one ready-made agent | You learn how an agent plans and acts on a real task. | It becomes the brain. |
| 3. Give it tools | You learn how an agent reaches files, the web and apps. | It becomes the hands. |
| 4. Add memory | You learn how to stop an agent forgetting you. | It becomes the shared memory. |
| 5. Schedule it | You learn how work starts with nobody at the keyboard. | It becomes the clock. |
| 6. Teach it skills | You learn how to make good output repeatable. | It becomes the playbook. |
| 7. Run several agents | You learn how to split work across roles. | It becomes the team and the task board. |
| 8. Build your own agent OS | You learn how to manage everything from one screen. | It becomes the dashboard. |
The order matters more than the speed.
Most people who quit started at step seven, with a team of agents and nothing for them to do.
Start with one agent doing one job, and add the next layer only when the last one works.
Quick-start: touch every step in one sitting
If you want to feel the whole path before you study any of it, this is the route I'd take on a fresh machine.
| Order | What you run or do | What you have afterwards |
|---|---|---|
| 1 | You install Hermes Desktop, or run curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash. | One agent is on your machine. |
| 2 | You run hermes setup --portal and pick a model. | The agent has a brain. |
| 3 | You give it one real task from your week. | You have seen the agent loop work. |
| 4 | You run hermes tools. | You know which tools are on. |
| 5 | You make a notes folder with an about-me note, a decisions log and a daily log. | The agent has a memory it can read and write. |
| 6 | You run hermes cron create "0 9 * * *" "Write my morning brief". | One job runs every morning. |
| 7 | You run hermes skills list and try one skill as a slash command. | You have used your first skill. |
| 8 | You run hermes profile create research. | You have a second agent with its own memory. |
| 9 | You run hermes dashboard. | A web dashboard opens on your own machine. |
You will not be good at any of these after one sitting.
You will know what each layer is for, and that makes every lesson afterwards easier to place.
I checked these commands against my own Hermes install on version 0.21.4.
Step 1: understand what an agent is
A chatbot answers a question, and an agent finishes a task.
Anthropic's engineering guide says agents are systems where the model dynamically directs its own process and tool usage.
It contrasts them with workflows, where models and tools follow predefined code paths.
In plain words, an agent has a brain, hands and a loop.
The brain is the model, the hands are the tools, and the loop lets it check its own work and try again.
An Agent OS is simply what you get when you give several of those agents one memory and one place to work.
That same Anthropic guide recommends the simplest solution possible, with complexity added only when it is needed.
I teach the same thing, which is why this path begins with one agent.
Step 2: use one ready-made agent for a real task
You do not need to build an agent to learn agents, because strong ones are free.
I start people on Hermes Agent, which is a free, open-source agent from Nous Research.
The official docs give you the Hermes Desktop installer or a one-line terminal command.
After that, hermes setup --portal logs you in and lets you choose a model.
Some of the models there are marked free, so check the label before you confirm.
My rule is free models first, because you can learn the method without a bill.
Free models get rate limited at busy times, so hermes fallback add gives you a second one.
If something breaks, hermes doctor is a built-in health check that explains the problem in plain words.
Now hand it one real task, such as a summary of a folder of your own notes.
In an Agent OS this agent becomes the brain, and I wrap mine in one profile per model.
🔥 Want the exact learning AI agents 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.
Step 3: give it tools
Tools are how an agent acts on the world.
Running hermes tools shows every toolset as enabled or disabled.
On my install it listed 26 built-in toolsets with 20 switched on.
The terminal, files, web and scheduling toolsets are on by default.
You add one with a command such as hermes tools enable browser.
Outside apps plug in through MCP, which the official site describes as an open-source standard for connecting AI applications to external systems.
You run hermes mcp catalog to browse and hermes mcp install to add one.
Tools are set per platform, so your phone bot can carry a shorter list than your desk.
In my Agent OS the terminal toolset comes first, because every other room depends on an agent that can actually run things.
The first mistake in my crash course is adding too many tools too soon, so add one and test it.
Step 4: add memory
Memory is the layer that turns a clever stranger into a colleague.
Hermes has a small built-in memory that persists between sessions.
The docs describe a MEMORY.md file limited to 2,200 characters and a USER.md file limited to 1,375 characters.
Past sessions are stored in a local database that the agent can search.
I add a notes vault on top, and I use the free Obsidian app for mine.
Three notes start it, which are an about-me note, a decisions log and a daily log.
One rule runs it, which is that every agent reads before it works and writes before it leaves.
Give the agent the real folder path, because I once said "this folder" on camera and it built the vault in its own workspace.
This is the most important layer of an Agent OS, because every agent you add later reads the same brain.
Seven of my agents read one vault.
When a new model comes out, I change a dropdown and the new brain inherits everything.
Step 5: schedule it
A scheduled job is a prompt that runs on a timetable and delivers its result to you.
This is the real command from my use cases guide.
hermes cron create "0 6 * * *" "Read my task list and send me a short to-do recap" --name "Daily 6AM To-Do Recap" --deliver telegram
The schedule also accepts plain forms such as every 2h.
You list your jobs with hermes cron list, and hermes pause is the emergency stop.
A job that calls a model uses tokens on every run, so start with one or two.
In an Agent OS this layer is the clock, and it is what makes the system feel alive.
My own morning radar is one scheduled run that sweeps for news at 6am and sends the top stories to my phone.
Step 6: teach it skills
A skill is a reusable instruction pack that the agent loads only when a job needs it.
In Hermes a skill is a folder with a SKILL.md file inside it.
You see yours with hermes skills list, and every installed skill becomes a slash command.
You can correct the agent once and have the correction saved as a skill.
The /learn command goes further and turns a guide or a book into a skill.
In my Agent OS the same skills work from the chat window, from scheduled jobs and from messaging apps.
That is the point of this layer, because the playbook is shared by every door into the system.
Hermes also tidies its own skills, and my crash course shows a skill unused for a month being flagged.
Step 7: run several agents
You add a second agent when the first has too many different jobs.
In Hermes you do that with profiles, such as hermes profile create research and hermes profile create builder.
Each profile has its own memory, its own skills and its own model.
In my Agent OS every profile shows up as a chip in the chat automatically.
Each chip is one employee persona.
Sub-agents are the second way to spread work, and the ticket board is the third.
You create a board with hermes kanban boards create my-builds.
Bigger jobs then become tickets instead of prompts, and a Hermes worker picks each one up.
In my Command Desk guide, three tickets were finished without me, which were a landing page, a content kanban app and an analytics dashboard.
My crash course warns against five profiles with zero workflows, and against trusting sub-agents unchecked.
Step 8: build your own agent OS
An agent OS is one dashboard where your agents live, share one memory and pick up tasks from one board.
You have built the memory and the board already, so the dashboard is the last piece.
The quick version is hermes dashboard, which opens a web dashboard on your own machine.
The full version is one you direct yourself.
You describe the dashboard to a coding agent, look at what it made and tell it what to change.
That loop of describing, reviewing and refining is how every feature in my own system was built.
Mine has seven agents wired in.
You do not write code for this, because your role is the director.
My AI Agent OS guide gives the exact prompts in the exact order, and my free video course on it is at the top of this page.
🔥 Want the exact learning AI agents 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.
What to do each week for four weeks
This is a plan and not a promise, so repeat any week that does not finish with a working layer.
| Week | Steps | What you do | What exists in your Agent OS at the end |
|---|---|---|---|
| 1 | 1 and 2 | You read one definition, install one agent and finish one real task each day. | The brain is working. |
| 2 | 3 and 4 | You check your tools, add one, and build the notes vault. | The hands and the memory are working. |
| 3 | 5 and 6 | You schedule one job and save one correction as a skill. | The clock and the playbook are working. |
| 4 | 7 and 8 | You add a second profile, open a ticket board and open the dashboard. | You have a small Agent OS that you can grow. |
The test at the end of each week is simple.
You ask whether the agent finished a real job this week that you would otherwise have done yourself.
A resource for every step
Every guide in the middle column is free on this site.
I checked every outside resource on its own page on 13 October 2026, and I have not taken the outside courses myself.
Two broader courses cover the theory behind the full path.
The Hugging Face AI Agents Course describes itself as a free course that takes you from beginner to expert.
The Agentic AI course from DeepLearning.AI is taught by Andrew Ng, and its page says you can audit it for free.
That course asks for intermediate Python skills, so it suits you if you want to build agents from code.
My free community is the AI Money Lab, which includes a free AI course and 1,000 AI agents.
My paid community is the AI Profit Boardroom, which includes the installable Agent OS and four coaching calls a week.
If you want the same path from other angles, read how to learn AI agents from scratch, how to learn AI agents for free and my best AI agent course list.
How long it realistically takes
It depends, and I will not invent a number.
These are the only timings I can source.
| Source | What it says about time |
|---|---|
| The Hermes docs | The install is a single command. |
| My tools guide | The tools quick-start is a ten-minute route. |
| My memory guide | The notes vault takes about an afternoon to do properly. |
| The Hugging Face AI Agents Course | Each chapter is designed for one week, at roughly three to four hours of work a week. |
None of those say when you will be good.
That depends on how many hours you give it and whether you have a real job for the agent.
People with a real job to hand over learn faster, because every session ends with something they can use.
Seven mistakes that slow people down
These are the seven mistakes from my Hermes crash course.
- You add too many tools too soon.
- You save everything to memory.
- You put a weak model on hard work.
- You turn every idea into a scheduled job.
- You trust sub-agents without checking them.
- You create five profiles and have zero workflows.
- You treat the agent like magic, when it only compounds if you feed it well.
The fix for all of them is to finish one job before you start the next.
Also On Our Network
- 🌐 The eight steps to learn AI agents, with every command
- 🌐 How a business owner learns AI agents around one real job
- 🌐 Every resource for learning AI agents, compared and rated
- 🌐 Which AI model to learn agents on, with live GoldieBench scores
- 🌐 My best AI agent course breakdown on aiprofitboardroom.com
FAQ
What is the best way to learn AI agents?
Learn in order. Understand what an agent is, use one ready-made agent for a real task, then add tools, memory, a schedule, skills and more agents, and finish by building your own agent OS. Each step leaves a working layer behind.
Do I need to code to learn AI agents?
No. The path on this page uses copy-paste commands and plain English prompts. You need Python only if you want to build agents from scratch with frameworks.
How long does it take to learn AI agents?
It depends on your hours and on whether you have a real job for the agent. Hugging Face says each chapter of its agents course is designed for one week at about three to four hours a week. Nobody can honestly promise a date for mastery.
What is an Agent OS?
An Agent OS is one dashboard where your AI agents live, share one memory and pick up tasks from one board. Julian Goldie's has seven agents wired in.
Can I learn AI agents for free?
Yes. Hermes Agent is free and open source, some models are free, and Julian Goldie's YouTube courses, the guides on agentos.guide and the AI Money Lab are free. The Hugging Face AI Agents Course is free as well.
Which agent should I start with?
Julian Goldie starts people on Hermes Agent, because it is free, open source and installs with one command. The same eight steps work with other agents.
Final word on how to learn AI agents
Install one agent tonight, give it one real job, and add one layer a week, because that is how to learn AI agents.
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert and the founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
My YouTube channel has 400,000+ subscribers, and I run Goldie Agency, a seven-figure SEO agency.
Related guides on agentos.guide
→ Hermes AI Agent: The Full Crash Course
→ How To Build Your Own AI Agent OS
→ The Best Hermes Agent Use Cases
Get the whole Agent OS
The installable system behind every guide on this site — dashboard, agents, memory and pipelines — updated as new tools land, with 3,400+ members and four coaching calls a week.
Join the AI Profit Boardroom →📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉