📖 Terms of this lesson — in simple words
AI — artificial intelligence. A program that understands text and responds meaningfully. The most famous today are ChatGPT and Claude.
Chatbot — answers questions. Asked → answered. Does nothing by itself.
AI agent — is an AI that can ACT. It doesn't just respond with text — it opens websites, reads emails, searches for info, writes code, gets things done. By itself.
Prompt — a task you give to AI in plain language. "Do X for me" — that's a prompt.
Claude Code — AI agent from Anthropic. Lives in your terminal (black window with commands). Writes code, reads files, runs programs — according to your words.
Vibe-coding — the work style of 2026. You describe what you want in words — AI writes the code. You don't learn syntax. You manage like a manager. You control like an architect.
❌ Before this lesson
- You think AI is just a chat for questions
- You don't understand how a program can do something "by itself"
- You believe that for AI you need to know Python and complex math
✅ After this lesson
- You understand how an agent differs from a chatbot
- You see what tasks you can assign to an agent instead of yourself
- You know that for your agent you only need: Claude Code and an idea
Chatbot vs AI agent — what's the difference?
Imagine two assistants.
The first is a librarian. You ask: "Who wrote "War and Peace"?" — he answers: "Tolstoy." Asked → got an answer. He does nothing more. This is a chatbot.
The second is your intern. You say: "Find 10 fresh news about crypto, select the most important ones, write a post on Telegram and publish it." The intern leaves, opens 10 websites, reads, selects, writes, clicks a button. After 15 minutes, brings it ready. This is an AI agent.
📚 Chatbot (old format)
- Responds with text
- Cannot open a website
- Cannot read a file
- Does not remember yesterday's conversation
- One step — one answer
- You copy and do it yourself
🤖 AI agent (new format)
- Acts — opens websites, files, programs
- Uses tools — browser, email, code
- Remembers previous steps
- Performs multi-step tasks
- Checks the result — redoes if bad
- You are the manager, the agent is the executor
💡 Main idea
Chatbot = interlocutor. AI agent = your subordinate. The difference is like between talking to a friend and hiring an employee.
How does an AI agent make decisions?
Inside the agent — a simple loop that repeats until the task is solved. Don't be scared, this is not code — this is logic:
Step 1 · Understood the task
You said what needed to be done
"Find 10 news about Bitcoin for today" — the agent understood that it needed to open news websites.
↓
Step 2 · Chose a tool
What will I use to do it?
The agent has "hands" — browser, files, code. It chooses "open website" (not "read file") because the news is on the internet.
↓
Step 3 · Performed the action
Opened, read, recorded
Visited coindesk.com, downloaded text, saved to memory. Moved to the next website.
↓
Step 4 · Checked the result
Did it turn out well?
"Did I collect 10 news items? Yes → I move on. No → I'll repeat the step with another website."
↓
Step 5 · Final
Done, reported to you
"Here are 10 news items, here's the post, published to your channel at 14:32."
This cycle — "think → choose tool → act → check" — is called agent loop. It was invented in 2023. Today, all smart AI systems are built on it. We will study it in its entirety in Module 3.
What can your agent actually do?
Not theory — our own agents that are working on the server right now and are sold to clients. Prices are real:
🤖 AI Chatbot for Website
from $149
Trained on your catalog, prices, FAQ. Answers clients 24/7, captures leads, transfers to Telegram.
guardlabs.online → automation ↗
A 3–7 step quiz that guides the client to purchase. AI calculates the score, recommends a product, collects email.
guardlabs.online → quiz ↗
📲 Telegram Lead-bot
from $79 + $19/month
Any form on the website → instantly to your Telegram. AI writes the first follow-up email for you if the client doesn't respond.
guardlabs.online → telegram ↗
🔍 Web Audit agent
from $9/month
Every 30 minutes checks your website: if it's down, if SSL is broken, if there's drift in localization. Sends an alert to Telegram.
guardlabs.online → web audit ↗
🛡️ Antifraud detector
from $49
AI looks at registrations, IP, behavior — distinguishes bots from real clients. Saved us from 99 fakes in 1 night.
guardlabs.online → antifraud ↗
Multilingual AI assistant for questions about life, habits, goals. Calibrates to you in 5 minutes.
askoracle.site ↗
💎 What this means for you
Each of these agents we sell from $9 to $1499. After the course, you will be able to make them yourself — for yourself or for order. The course costs $199. To recoup it — you need 1 large order (NEXUS Bot DFY $499–$1499) or 4–6 simple agents (Lead-bot $79+$19/month, Quiz $99). After that — each subsequent order is pure profit.
💡 Rule for choosing a task for an agent
A good task for an agent is a routine that repeats. Every morning, gather news. Every single day, check email. Set it up once — save several hours a week for the rest of your life.
Why should you learn this now?
As recently as 2022, to build such an agent, you needed a programmer with 2-3 years of experience. Python, asynchronous programming, API, frameworks — the entry barrier was high.
In 2024, Claude Code appeared — an AI agent that you control through plain English. You write: "Make me an email assistant that reads Gmail and replies to simple emails" — Claude Code writes the code, installs libraries, runs the program, checks that it works. All that's left for you is to check that the logic is correct.
Before (without AI)
You learn Python — 6 months. You learn API — 2 months. You learn deployment — 1 month. Total: 9+ months until the first working agent.
Now (with Claude Code)
You + Claude Code. You describe the task in plain English. AI writes the code. You check the logic. First agent in 1–2 days. This is the skill of 2026.
What we will do in this course
- Module 1 — we install Claude Code, create the first prompt, get the first assistant agent.
- Module 2 — we teach the agent to work with the internet and files. It begins to "see" the world.
- Module 3 — we give the agent memory. It remembers what happened yesterday. Plans for tomorrow.
- Module 4 — connection to the outside world: email, Telegram, databases, your services.
- Module 5 — we teach the agent to think in multiple steps. Complex tasks → plan → execution.
- Module 6 — we deploy the agent to the server. It works 24/7 without you. You — in a lounge chair.
- Modules 7-10 — your first working turnkey projects: news agent, content agent, lead hunter, code assistant.
🎯 Main takeaway from this lesson
AI agent = your subordinate who can act, not just respond. To build your own — you need Claude Code and an idea. You do NOT need to program yourself — you manage AI like a manager. In the next lesson, we will install Claude Code on your computer and make the first launch.