Automation
What Is n8n? A Practical Guide to Workflow Automation
9 min read
n8n is an open-source workflow automation platform. You connect apps, databases, and APIs together as a series of nodes on a canvas, and n8n runs that workflow on a schedule, on a webhook, or on demand. The short answer to 'what is n8n' is: it is the automation layer between the tools you already pay for, and unlike most of its competitors you can host it yourself.
A workflow in n8n is a graph. A trigger node starts it — a webhook, a cron schedule, a new row in a database, an incoming email. Downstream nodes then transform, branch, filter, and write. Each node passes structured items to the next, so a workflow that processes one record processes a thousand with no change in design.
The feature that separates n8n from purely visual tools is the Code node. When a transformation is easier to express in JavaScript or Python than in a chain of UI steps, you write it inline. You are never blocked by the absence of a pre-built integration either: the HTTP Request node speaks to any REST API, and credentials are stored and reused like any first-class connector.
How does n8n compare to Zapier and Make? Zapier is the easiest to start with and the most expensive at volume, because it prices per task executed. Make sits in the middle with a stronger visual builder and cheaper operations. n8n is the most technical of the three and the cheapest at scale — self-hosted, your only cost is the server, and execution volume is effectively free. If your automation touches sensitive data that cannot leave your infrastructure, self-hosting is not a cost decision at all, it is the only viable option.
The trade-off is real. Self-hosted n8n is software you now operate: upgrades, backups, queue workers, and monitoring are yours. n8n Cloud removes that burden and still costs less per execution than Zapier at most volumes. We generally recommend Cloud until either data residency or execution volume makes self-hosting worth the operational overhead.
Where n8n earns its keep: AI pipelines. Its LangChain nodes let a workflow call a model, run retrieval against a vector store, use tools, and hand structured output to the next step — with retries, error branches, and logging around it. That is the difference between a demo script and something a business can depend on.
Typical deployments we build look like this. Lead routing: a form submission enriches against a data provider, scores against your criteria, writes to the CRM, and alerts the right person in Slack. Document processing: an inbound invoice is parsed by a model, validated against purchase orders, and queued for human approval only when it fails a check. Content operations: a single source publishes to a CMS, a newsletter, and three social platforms with per-channel formatting. Internal reporting: overnight pulls from several systems reconcile into one dashboard table.
Before you build anything, measure the baseline. How long does the task take today, how often does it run, and what is the current error rate? Those three numbers tell you whether the workflow is worth building and what accuracy it has to hit. Automation that halves processing time but doubles review time has moved the work, not removed it.
Three practices keep n8n workflows from becoming liabilities. Put an error workflow on everything, so failures alert a human instead of vanishing. Keep workflows small and call sub-workflows, because a fifty-node canvas is unmaintainable. Version your workflows in Git — n8n exports JSON, and treating that as code means you can review changes rather than discover them.
n8n is the right choice when you have real volume, technical capacity in-house or on retainer, integrations that no off-the-shelf connector covers, or data that must stay on your own infrastructure. It is the wrong choice when you need three simple automations and nobody to maintain them; buy the simpler tool and spend the time elsewhere.
If you are weighing this up for a specific process, we will map the workflow, estimate the return against your baseline, and tell you honestly whether n8n is the right layer for it.
