n8n
Workflow automation and AI agents
The automation canvas that rewards real engineering thinking, near-free if you can run the ops yourself.
Technical teams building AI agents and workflows who want a visual canvas but can drop into JavaScript or Python the moment a node runs out of road.
Not truly no-code. The gap between looks-easy and works-in-production is wide, and self-hosting is genuine ops work you own forever.
Reach for it when someone technical owns it and your workflows run past ten steps. If you need hands-off no-code, it will fight you.
What you're actually getting
n8n is a node-based automation canvas for building workflows and AI agents you can watch run. It ships 500+ integration nodes, and anything without a native node you wire up with a raw HTTP or GraphQL request.
The difference from Zapier and Make is the code escape hatch. Write JavaScript or Python inline anywhere in a workflow, so you never hit a ceiling when logic gets messy.
Deploy it your way. Run n8n Cloud hosted with managed AI credits, or self-host the free Community Edition on your own server and pay nothing but hosting.
The tools I run in n8n are connected to dozens of APIs, each doing one specific thing. What n8n does for me is take information and then send and process it through those APIs. I haven't tried doing it via Claude, but from my experience, when it has to remember dozens of variables and different content types, things get lost. Maybe I'm old school, though half a year with this tool is already old school.
Where it earns its keep
Cost is the loudest win. Per-execution pricing means n8n beats Make once a workflow passes ten steps, and self-hosting on cheap rented server space makes every run essentially free.
It earns its keep on complex, branching automation and AI agent builds. The HTTP node, JS execution, and webhook handling push you toward real engineering, so it stays interesting long after simple no-code tools stall.
Governance is there when you grow into it: version control, separate test and live environments, permission controls, SSO, and audit logs on the higher tiers.
I added a self hosted $5/mo VPS with n8n, to run some workflows that are expensive … a single scenario run on Make would cost several dollars! And self hosting means every n8n execution is completely free so long as the VPS can handle the computation cost.
Where it'll bite you
The no-code label oversells it. For non-technical users, credentials break, APIs change, and hosting fails, and a supposedly quick automation turns into two or three days of debugging. Put a technical owner on it before you depend on it.
Self-hosting is not free labor. Server rental is cheap, but you stay on the hook for updates, backups, and breakage. Budget the maintenance time honestly.
Custom code carries a licensing catch. Ship a public commercial API built on n8n custom code and you can owe a commercial license, and users report noticeable latency added per custom-code node, so it is a weak fit as a high-performance backend at scale.
n8n's downside is complex maintenance. At a certain scale you need a queue-management system, a relatively complex database (PostgreSQL) and Docker, all of that just to do a hello world. There are several workflow engines that make a great alternative to n8n if all you want is to run code on a schedule or event-driven.
What it costs, really
Cloud runs on monthly executions, where one execution is a full workflow run regardless of steps. Prices are in euros: Starter €20/mo, Pro €50/mo, and Business €667/mo for self-hosted with SSO and Git. Enterprise is a custom quote, and annual billing saves about 17 percent.
The genuinely free path is the self-hosted Community Edition from GitHub, where you pay only for the server. Business and Enterprise features on self-hosted still need a paid license key.
pricing honestly depends. different n8n plans have limits on how many active workflows you can have live, make doesn't. If your workflows are less than ~8-10 steps, Make is usually cheaper. Self hosting n8n is cheap, but it's not as easy as it might sound, and you have to be on top of updates and changes.
N8n has 2 main problems: Commercial Licensing and Scalability … if you start custom coding (ex. in Python) with n8n, and then launch it as public commercial API you have to pay for a commercial license (up to $10k+) whilst running a suboptimal backend. The scalability issue is because n8n introduced noticable wait periods for each custom code node.
Same job, other tools
Also filed under Workflow & agent infrastructure:
