🌱 New here?
You’ll see names like coding assistant, n8n, MCP, skill, workflow e agent. Each one is explained in a “What is X?” box the first time it appears. Don’t memorize them: the goal is to understand the role for each piece. Think of a team: there’s the brain (assistant), the assembly line (n8n), the outlets that connect external apps (MCP), the procedure manuals (skills), and the autonomous employee (agent).
See the entire ecosystem at once. At the center is the coding assistant — the brain that reads, writes, and decides. Around it are the pieces it controls or uses: the visual builder (n8n), the protocol that connects external tools (MCP), reusable instructions (skills/workflows), and the final result, which can run on your machine or on its own in the cloud.
Illustrative diagram — the assistant (cyan) is the coordinator at the center; around it are the pieces it controls or uses; on the right is where the result can run. You’ll get to know each one next.
🧠 The code assistant
What it is
O coding assistant is the ecosystem’s brain: a AI that lives inside your text editor and can do four concrete things — read your project files, write and edit code, run commands in the terminal and connect external tools via MCP. It’s how you talk in natural language; everything else (n8n, skills, agents) is either directed by it or feeds its work.
What is it?Text editor (code editor) — the program where the assistant lives and where the project files are visible (chat panel, file explorer on the left, terminal at the bottom). Terminal — the window where you type commands for the computer; the assistant can use it for you.
One important principle right here: one folder per project. You open only that project’s dedicated folder—never a root folder containing multiple projects. That way, the assistant sees only what matters and doesn’t get confused or mix contexts. There’s also permission modes (ask before editing, edit automatically, plan mode…) that control how much the assistant acts on its own.
Why learn
Understanding the assistant as coordinator changes everything: you stop looking for "which button do I press?" and start thinking "what outcome should I describe?" It is the piece that turns your sentence into real files, commands, and connections.
- •Read and write: can see your files and create/edit whatever is needed.
- •Runs commands: installs things, runs scripts — without you having to memorize commands.
- •Connects the rest: via MCP, gets access to external apps and services.
Key concepts
Coordinates all the pieces.
Read, write, run, connect.
Open only the dedicated folder.
Control how much it acts.
🔗 n8n — the visual builder
What it is
O n8n is a builder visual of workflows (automation workflows) of the low-code/no-code kind: instead of writing code, you connect nodes (nodes) on a screen, and each node does one thing—read an email, filter, call an API, write to a spreadsheet. By connecting the nodes, you create automations that connect apps, data, AI, and external services.
What is it?Workflow (flow) — a sequence of automated steps, from trigger to action. Node (node) — each box in the workflow, with a single job. Low/no-code — build with little or no coding by connecting visual blocks.
Building n8n by hand is powerful, but labor-intensive and error-prone—a single incorrect setting in a node is hard to find. That’s exactly where the assistant comes in: connected to your n8n, it reads, creates, modifies, tests, and debugs the workflows for you. n8n becomes the deterministic “assembly line”; the assistant becomes the one who builds and fixes it.
Illustrative diagram — each node has a job and passes the result to the next. The assistant connects and adjusts these nodes for you.
Why learn
n8n is where many automations “live.” Knowing that it is deterministic (the same input produces the same output) and visual helps explain why it’s reliable for repetitive tasks—and why combining n8n (the predictable pipeline) with AI (the flexible brain) is so powerful.
Key concepts
Connects blocks on one screen.
Each one does one thing.
Email, spreadsheets, APIs, AI.
Creates, tests, and debugs for you.
🔌 MCP — the tools protocol
What it is
MCP (Model Context Protocol, "model context protocol") is the standard that provides external tools to the assistant. Think of a universal outlet: a single connection (a MCP server) exposes a whole set of actions. For example, an email server exposes “send,” “draft,” and “search messages” — and the assistant decides on its own which action to call, when, and with which parameters.
What is it?Protocol — a shared “language” for two things to communicate. MCP server — the small program that connects the assistant to an external service (email, database, web scraper…). Tool — each action that server offers. API — a service’s “front door” for other programs to communicate with it.
Why learn
Without MCP, the assistant only knows what’s inside the editor. With MCP, it can reach the world: scrape the web, read a database, create documents, send messages. For example, MCP lets you connect the assistant to your n8n so it can build workflows there. Discovering and installing servers (through directories or repositories) opens up a wider range of tasks.
✓ What the MCP does
- ✓Exposes many tools through a single connection.
- ✓The assistant chooses the action and parameters on its own.
- ✓Connects categories: web, databases, documents, communication.
✗ What to avoid
- ✗Paste the API key into the chat (it goes in the
.env, not in the conversation). - ✗Leave many servers on without using them (uses up context).
- ✗Configuring it globally when it should be project-level.
Key concepts
One connection, many actions.
Connects the assistant to the service.
The available actions.
It chooses which one to use.
📄 Skills & workflows
What it is
Workflows e skills are basically the same thing: reusable instructions written in markdown (plain text with headings, lists, and bold). They work like a standard operating procedure (SOP) manual that tells the assistant what to do, step by step. The only difference is where they live and what they’re called: a skill lives in a special folder and becomes a quick command (slash command, e.g.: /rascunhar-email) that you invoke on demand.
What is it?Markdown — a simple way to format text (with # for titles, - for lists). SOP — “standard operating procedure,” a manual for how to do a task. Slash command — a slash command (/) that triggers a skill.
Why learn
Skills save you from re-explaining the same task every time. They provide consistency (same process every time), savings (load only when called, so they don't weigh down the context unnecessarily), and sharing (the team uses the same one). And since a skill can call tools (scripts that do the work), the instruction system + tools remain intact underneath.
🎯 Practical tip
Use a skill when the task repeats, requires consistency, or has steps that are easy to forget. For a one-off task, or one so simple it fits in a sentence, it's better to just type — a skill would be overkill.
Key concepts
Plain-text instructions.
Don't re-explain every time.
Invokes on demand.
Can trigger scripts.
🤖 Agents — when the system decides
What it is
A agent is what emerges when the system stops just “executing fixed steps” and starts to decide on its own. In traditional automation, you wire up each step and fix it when something breaks. In an agentic workflow, you say what result what you want, and the agent reasons, adapts, asks clarifying questions, and even self-corrects when something fails.
What is it?Agentic — behavior of something that reasons and chooses a path instead of following a fixed script. Self-healing (self-healing) — when the agent makes a mistake, diagnoses it, and fixes it on its own during the session. Deterministic × nondeterministic — traditional automation is predictable; the agent is flexible and can vary.
Illustrative diagram — the difference is in the bottom line: the agent doesn't get stuck on the error; it diagnoses and keeps going until it reaches the goal.
Why learn
Understanding what an agent is prevents two misconceptions: thinking that all AI "makes decisions on its own" (not always), and thinking that fixed automation is "dumb" (it’s deliberately reliable). The analogy helps: hiring an experienced developer (agent) is different from following a recipe to the letter (traditional automation). You choose the right tool for each task.
Key concepts
Reasons through the path.
Ask questions, adjust.
Gets it wrong, diagnoses, fixes.
You state the “what.”
🖥️☁️ Local vs hosted
What it is
The final piece is a choice: run locally or run hosted. Local = on your machine: you open the assistant, start the task, and it runs there while you're present. Hosted = in the cloud: the code is placed in a service that runs it alone, on the internet, at a scheduled time, or in response to an event—with no one to press the button.
What is it?Host (deploy) — run the code on a server on the internet, not on your machine. Cron (scheduling) — a clock that triggers the task at a set time (e.g., 7 a.m. every day). Webhook — a “digital doorbell”: a URL that wakes up the task when it receives data.
Why learn
Local is great for learning and building: you see everything happen and develop intuition. But human-triggered automation only scales as fast as someone can show up. Hosting is what makes automation work while you sleep. The price is the agentic gap: in the cloud, the assistant isn’t watching itself to heal, so reliability (logs, error handling, alerts) is your responsibility from the start.
🖥️ Local — your machine
- ✓You trigger it and monitor it in real time.
- ✓The assistant heals itself on the spot.
- ✓Ideal for learning and prototyping.
☁️ Hosted — the cloud
- ✓Runs on its own: on a schedule (cron) or in response to an event (webhook).
- ✓Scale without you being there.
- !Requires logs, error handling, and alerts (agentic gap).
Key concepts
On your machine, with you.
In the cloud, on its own.
By schedule or by event.
No self-healing in the cloud.
In the ecosystem, who is the coordinator that reads files, writes, runs commands, and triggers the other components?
📌 Module Summary
Next Module:
1.3 — Essential Glossary (Part 1: AI, Claude Code, n8n)