What is n8n, and when is it the right tool?

n8n is a workflow automation tool. You connect nodes on a canvas, each node does one thing, and data flows from one to the next. That description also fits Zapier and Make, so the useful question is not what n8n is but where it differs, and the answer comes down to two things: you can host it yourself, and you can drop into code whenever the visual tools run out.

What a workflow actually is

Every workflow starts with a trigger: a schedule, an incoming webhook, a new row in a sheet. After that it is a chain of operations. Fetch something, reshape it, make a decision, write it somewhere. The canvas is a readable version of a script you would otherwise write and then have to maintain alone.

The reason this matters is maintenance rather than creation. Anyone can write the script. Six months later, when it breaks and you have forgotten how it worked, a canvas with twelve labelled nodes is considerably easier to repair than two hundred lines of someone else’s Python.

Where it beats a hosted no-code tool

• Pricing is per execution on most hosted tools. Self-hosted n8n has no execution ceiling, which changes the economics completely once you are running thousands of jobs a month.
• Data stays on your infrastructure, which settles a lot of questions when the data is personal or under contract.
• A Code node takes JavaScript or Python, so the moment the visual nodes cannot express something, you write six lines instead of abandoning the workflow.
• Workflows are JSON, so they live in version control and move between environments.

Where it does not

Self-hosting is not free in the sense people usually mean. Somebody updates it, watches it, and fixes it at an inconvenient hour. If that person is you and your time is worth more than the subscription, a hosted plan is the cheaper option however the pricing page reads.

The node library is also smaller than the largest hosted competitors. For common services this is irrelevant. For a niche CRM, check before committing, because building an integration yourself is a different project from configuring one.

When to reach for code instead

Automation tools are at their best gluing systems together. They are at their worst holding complex business logic. If a workflow has grown past roughly thirty nodes and most of them are conditionals, that logic wants to be a small service with tests, called by a workflow that stays simple.

The signal to watch for is debugging time. When finding a bug takes longer on the canvas than it would in an editor, the canvas has stopped earning its place.

A reasonable first project

Pick something that runs on a schedule, fails safely, and nobody depends on this week. A report that pulls numbers and emails a summary is close to ideal: the trigger is simple, the output is visible, and if it breaks on a Tuesday the consequence is a missing email rather than a missing invoice.

Avoid starting with anything that writes to a billing system. Write-access automations deserve to be your fourth workflow, not your first.