When you open a visual automation tool, you see rectangles joined by lines. Each rectangle is a node. It is the unit you will build with, and the minimum vocabulary needed to follow any tutorial.
Definition
A node is a single step of a Workflow: it receives data from the previous step, performs an action, and passes its result to the next.
The rule that makes everything comprehensible fits in one sentence: a node does one thing and one thing only. Send an email. Filter a list. Call a service. That granularity is what lets you see where it breaks when it breaks.
The families of nodes
| Family | What it does | Example |
|---|---|---|
| Trigger | Starts the workflow | Receiving a Webhook |
| Action | Acts on an outside service | Create a client record |
| Logic | Directs the flow | Condition, branch, loop |
| Transformation | Reshapes the data | Rename fields, merge |
Beginners focus on action nodes, which are the most visible. In practice it is transformation nodes that take up half of a real workflow: two services rarely speak the same language, and something has to translate between them.
A good habit from day one: rename your nodes. "Create client record" still reads six months later, "HTTP Request 3" does not read at all. That is time saved on every future debugging session.
Best practices
Look at the data coming out of each node. It is debugging reflex number one. Nine times out of ten, a workflow failing at step five received something unexpected at step four.
Do not chain too far. Past twenty or so nodes, split into several workflows that call each other. Readability beats the elegance of one large diagram.
Add an error node. Most tools let you branch a specific path on failure. That is where you put the alert that will warn you.
Test node by node. Run it, look at the output, add the next. Building ten nodes at once and then running is the surest way to search for a long time.
Frequently asked questions
What if the service I need has no dedicated node?
Use the generic API call node, present in every tool. It requires reading the service documentation, but it lets you connect to almost anything, including what the vendor did not anticipate.
Can a node handle several items at once?
Yes, most run once per item received. That is very convenient, and also a source of nasty surprises: a node receiving a thousand items runs a thousand times, and consumes accordingly. See Batch processing.
Can you write code in a node?
Yes, every tool offers a code node for cases the standard blocks do not cover. It is useful, but each code node is one more place to maintain: keep it for what genuinely cannot be done otherwise.
How do you learn to assemble them?
By building a real workflow rather than reading through the list of available nodes. Our n8n course works that way, introducing each family at the point where the case makes it necessary.