How To Learn AI Agents From Scratch: My 14-Day Plan

How to learn AI agents from scratch comes down to one rule, which is that you set up one agent the easy way and give it one small job every day for 14 days.

I'm Julian Goldie, and I've written this plan for someone who has never opened a terminal and has never seen an API key.

You won't write any code in these two weeks, and you won't pay for a model either.

Here is the whole path in six lines, so you can see where you're going before you start.

On day 1, you learn what an AI agent is in plain words.
On day 2, you install one agent as a normal desktop app.
On day 3, you give it three tiny tasks and watch it work.
On days 4 to 7, you learn what a model, a provider and an API key are, and you move onto a free model.
On days 8 to 10, you build your first automation, which is a job that runs on a schedule without you.
On days 11 to 14, you save your first skill, get unstuck with other people and decide what to learn next.

The agent I use in this walkthrough is Hermes Agent, because it's free, it's open source and it now installs like any other app.

The 14-Day Plan To Learn AI Agents From Scratch At A Glance

This table is the whole plan on one screen, and each day should take you 20 to 40 minutes.

Day	What you do	What you have at the end
1	You learn what an agent is and pick one job to hand over.	You have a one-line goal.
2	You install Hermes Desktop and send a first message.	You have a working agent.
3	You run three tiny tasks.	You've watched an agent plan, act and report.
4 and 5	You learn the words model, provider and API key, and you check you're on a free model.	You know your agent costs nothing.
6 and 7	You write an "about me" note and join a free community.	You get better answers and a place to ask questions.
8 to 10	You hand over one real job and put it on a schedule.	You have your first automation.
11 and 12	You save that job as a skill and reuse it.	You have your first skill.
13 and 14	You practise getting unstuck and pick one next step.	You have a plan for month one.

I've kept every day small on purpose, because beginners quit when day two takes four hours.

Day 1: What An AI Agent Is, In Plain Words

A chatbot answers your question, and an AI agent goes and does the job.

That's the only difference you need on day one.

When you ask a chatbot to plan your week, it writes you a plan and stops.

When you ask an agent to plan your week, it can read your notes, check a web page, write a file and tell you what it did.

An agent does that by running a simple loop, and the loop has four parts.

It reads your goal.
It makes a short plan.
It uses a tool, such as web search or a file on your computer.
It checks the result and decides whether to keep going.

That loop is the whole idea, and everything else you'll learn is a detail on top of it.

Your only homework today is to write down one boring job you'd love to hand over.

A good first job is small, it happens often and nothing breaks if the agent gets it wrong.

"Summarise the three newsletters I never read" is a good first job, and "run my whole business" is not.

Day 2: Set Up One Agent The Easy Way

Today you install one agent, and you don't touch a terminal to do it.

Hermes Agent is a free, open-source AI agent from Nous Research, and it has an official app called Hermes Desktop.

Hermes Desktop installs like any other program, which is why I send beginners there first.

Go to the Hermes Desktop page and download the package for your computer.
On a Mac, open the DMG file and drag Hermes into your Applications folder.
On Windows, open the download and let Windows App Installer finish the job.
Open the app and sign in to Nous Portal when it asks you to.
Pick a model from the list, and choose one that is marked free.
Type "test" into the chat and wait for a reply.

The Mac package is for Apple Silicon Macs only, so an older Intel Mac won't run it.

If that's you, use one of the backup routes below.

Once you get a reply to "test", you're done for the day.

A working agent and a closed laptop is a perfect day two.

The two backup routes if Hermes Desktop doesn't suit you

The first backup is Hermes Cloud, which is the hosted version that Nous Research runs for you.

You go to portal.nousresearch.com/cloud, sign in, create an agent, pick a free model and send it a test message.

Nothing runs on your computer, so it suits an old laptop.

Hermes Cloud is a paid product, and when I checked in October 2026 an agent needed a $2 minimum in credits or an active Nous subscription to deploy, so check the current price on that page first.

The second backup is the Ollama route from my one-click video.

You install the free Ollama app from ollama.com, make sure it's open, copy the Hermes launch command from a model page and paste it into your terminal.

That route uses the terminal for one paste, and it takes about 20 seconds once Ollama is running.

My one-click Hermes setup walkthrough covers that route line by line.

My best Hermes Agent setup ranking compares all the setups once you're ready for more.

🔥 Want someone to watch you do day two and fix what breaks? Inside the AI Profit Boardroom, there's a full Hermes Agent course plus four coaching calls a week where you can share your screen and get your setup sorted. You'll be learning alongside 3,400+ members who started exactly where you are. → Get access here

Day 3: Your First Three Tiny Tasks

Today you give your agent three jobs that each take under five minutes.

The point isn't the result, because the point is watching the loop from day one happen in front of you.

Tiny task 1: Summarise one web page

Paste a link to any article and type "Read this page and give me the five most useful points in plain English."

Watch for the moment the agent says it's fetching the page, because that's the agent using a tool.

Tiny task 2: Make one file

Type "Write me a packing list for a three-day work trip and save it as a file on my computer."

If the app asks for permission before it writes, read what it's asking before you click yes.

Then go and find the file, because seeing a real file on your own machine is when agents stop feeling like chat.

Tiny task 3: Research and report

Type "Find three free tools for making invoices, compare them in a small table and tell me which one you'd pick."

This task has several steps, so you'll see the agent plan, search, read and write.

If a task fails, that's normal on a free model.

Ask the same thing again in simpler words, and break a big request into two smaller ones.

Days 4 And 5: Models, Providers, API Keys And Using A Free Model

You've been using three things for two days without knowing their names, so let's name them.

Word	What it means in plain English	What it looks like for you
Model	It's the brain that does the thinking.	It's the name you picked from the list on day two.
Provider	It's the company that runs the brain and lets your agent use it.	It's Nous Portal in this plan.
API key	It's a long password that lets your agent use a provider.	You haven't needed one, because you signed in instead.
Token	It's a small piece of a word, and providers count usage in tokens.	It's why free models have limits.
Rate limit	It's a cap on how much you can use in a set time.	It's the "try again later" message on a busy day.

The agent is the worker, the model is the brain and the provider is the place the brain lives.

Hermes Agent doesn't come with one fixed model, which means you can swap the brain without reinstalling anything.

On day five, open the model settings and check three things.

Check that the model you're using is marked free.
Check for a twin, because some models appear in both a free and a paid version and you want the entry marked free.
Pick a second free model as your backup for busy times.

Free lists change, so don't get attached to one name.

When I recorded my Solar Mini 4 video, the Nous Portal free list included Upstage Solar Mini 4, NVIDIA Nemotron, Poolside Laguna XS 2.1 and a few others.

If a free model you expected is missing, refresh the model list, because the list is cached.

My Solar Mini 4 Hermes guide shows that free-model setup click by click.

If you ever do get an API key, treat it like a bank card number and never paste it into a chat, a screenshot or a public post.

Days 6 And 7: Give It Context, Then Find Your People

On day six, you fix the most common beginner complaint, which is that the agent sounds generic.

It sounds generic because it knows nothing about you.

Write a short note with five lines about who you are, what you sell or do, who you serve, how you like to write and what you're working on this month.

Save it as a plain text file and tell the agent to read that file before it starts a job.

On day seven, you don't build anything.

You look back at the week and write down which task worked best and which one failed.

Then you join one free community, so you're not learning alone.

My free one is AI Money Lab, and its about page lists free AI courses, 50+ free AI tools and 1,000+ n8n workflows.

Post one line saying what you built this week and what confused you.

Days 8 To 10: Your First Automation

An automation is a job that runs on a schedule without you asking.

On day eight, do the job by hand first.

Pick the boring job you wrote down on day one and ask the agent to do it once while you watch.

A good example is "Read these three pages, pull out anything about my industry and write me a ten-line morning brief."

Fix the wording until the result is something you'd actually read.

On day nine, put that job on a timer.

You can type "Run this every weekday at 7am" in plain English, or you can use Automation Blueprints, which let you fill in a recipe instead of typing a schedule.

Under the bonnet, Hermes calls this a cron job, and the terminal command is hermes cron create, but you don't need the command to use the feature.

On day ten, read what arrived and improve the instructions.

Your first automatic result will be about 70% right, and that's fine.

Add one line to the instructions for each thing that annoyed you.

There's one honest limit you should know.

An agent on your laptop only runs while the laptop is on and Hermes is open, so a 7am job won't fire if the lid is closed.

Hermes Cloud or a small server solves that when you're ready.

My best Hermes Agent use cases post lists the automations I'd try next.

🔥 Want my real automations instead of guessing your first one? The AI Profit Boardroom includes my installable Agent OS, step-by-step courses on Hermes Agent and other agents, and four coaching calls a week where you can ask what to automate first. → Join the AI Profit Boardroom

Days 11 And 12: Your First Skill

A skill is a saved set of instructions that teaches your agent how to do one kind of job properly.

Without a skill, your agent works out the job from scratch every time.

With a skill, it follows the steps that already worked.

In Hermes, a skill is just a folder with a text file called SKILL.md inside it, so there's still no code involved.

On day eleven, run your day-eight job once more in a fresh chat.

When it finishes, type /learn and tell Hermes to turn the workflow it just ran into a skill.

The /learn command can also build a skill from a folder, a web page or a PDF.

You can ask in plain English as well, with something like "Save the steps you just followed as a skill called morning-brief."

On day twelve, start a new chat and use the skill on a different input.

If the result is good, you now own a repeatable process.

If it's off, tell the agent what to change and ask it to update the skill.

Hermes also ships with a library of bundled skills, and my best Hermes Agent skills ranking covers the ones worth turning on first.

Days 13 And 14: Get Unstuck, Then Choose What Comes Next

Everyone gets stuck, so day thirteen is for practising how to get help.

Here's the order I'd work through.

Ask the agent itself, because "what went wrong on that last step?" works more often than you'd expect.
Run the built-in health check, which is hermes doctor if you're comfortable in a terminal.
Search the community you joined for the exact error message.
Post your question with a screenshot, what you expected and what happened instead.
Bring the problem to a live call if you want someone to look at your screen.

A good question gets a fast answer, and "it doesn't work" isn't a good question.

On day fourteen, write three lines.

Write down what your agent does for you now, what you'd like it to do next and what confused you most.

Then pick one next step, and make it a second job for the same agent and not a second tool.

What You Do NOT Need To Learn Yet

Most beginners stall because they try to learn the whole field in week one.

You can safely ignore all of this for now.

Topic	Why you can skip it for now	When it starts to matter
Coding in Python	Everything in this plan is typed in plain English.	It matters when you want to build custom tools.
Fine-tuning	A good prompt and a skill get you most of the way.	It rarely matters for business users.
Vector databases and RAG	A notes file gives your agent enough memory to start.	It matters when you have thousands of documents.
Agent frameworks	You're using a finished agent, so you don't need to build one.	It matters if you become a developer.
Multi-agent teams	One agent with one job teaches you more.	It matters after your first few automations.
Running models on your own machine	Free hosted models are easier on day one.	It matters when privacy or cost becomes a concern.
Servers and Docker	Your laptop is enough for the first month.	It matters when you need the agent on all day.

When you do want the code side, two free courses are worth bookmarking.

The Hugging Face AI Agents Course describes itself as a free course that goes from beginner to expert.

Microsoft's AI Agents for Beginners is a free set of lessons on GitHub about the fundamentals of building agents.

I haven't taken either of them, so I'm pointing you to what their own pages say and not giving you a review.

Both are aimed at people who write code, which is exactly why they belong after these 14 days.

A Plain-English Glossary For Complete Beginners

The table on days four and five covers model, provider, API key, token and rate limit, and these are the other words you'll meet.

Agent: An agent is an AI that takes actions to finish a goal.
Tool: A tool is something the agent can use, such as web search, a browser or your files.
Terminal: A terminal is a window where you type commands, and you don't need one for this plan.
Profile: A profile is a separate copy of your agent with its own model, settings and memory.
Memory: Memory is what the agent keeps between chats, such as your about-me note.
Skill: A skill is a saved set of instructions for one kind of job.
Automation: An automation is a job that runs on a schedule.
Agent OS: An Agent OS is a dashboard that runs several agents under one shared memory.

Also On Our Network

FAQ: How To Learn AI Agents From Scratch

How do I learn AI agents from scratch with no experience?

Install one agent as a desktop app, give it three tiny tasks and then give it one small job a day for two weeks.

Hermes Desktop is the route I'd start with, because it installs like a normal app and can run on a free model.

You don't need to code, and you don't need a paid plan to begin.

Do I need to know how to code to learn AI agents?

No, you don't need to code to start.

Everything in this 14-day plan is typed in plain English.

Coding becomes useful later if you want to build your own tools.

What is an API key, and do I need one?

An API key is a long password that lets your agent use a model from a provider.

You don't need one for this plan, because Hermes Desktop lets you sign in to Nous Portal instead.

If you get one later, keep it private and never paste it into a chat or a screenshot.

Can I learn AI agents for free?

Yes, because Hermes Agent is free and open source, and you can choose a model that is marked free on Nous Portal.

Free models have rate limits, so keep a second free model as a backup.

What should I do when my agent gets something wrong?

Ask it what went wrong, make the request smaller and try again in simpler words.

If you're still stuck, post the error in a community with a screenshot, or bring it to a live coaching call.

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

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

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

I started Goldie Agency on my own, and it is now a 7-figure SEO and link-building agency with 70+ people.
I've grown my YouTube channel to 400,000+ subscribers.
I wrote "Link Building Mastery".
My Udemy courses have taught 29,000+ students.

→ Get my best AI training inside the AI Profit Boardroom

Install one agent, give it one small job a day for 14 days, and you'll know how to learn AI agents from scratch.