n8n self-hosted vs cloud: what small teams need to know is really a question about operational risk, because the workflow builder is identical on both sides. n8n can run on a server you manage yourself or as a hosted service. What changes is who patches it, who notices when it stops, and where the data moving through your workflows actually sits. For a small team in Singapore, the decision usually turns on three things: whether someone on staff can genuinely own a server, how much data residency matters under PDPA, and how much you expect your automations to grow in 2026.
Total cost is mostly maintenance effort
Self-hosting looks cheap because the visible line item is a server. The invisible line item is the hours. With self-hosting, your workflow data stays on your own infrastructure, and you take on updates, backups, server security and uptime monitoring. Each of those is recurring work, not a setup step you finish once. A backup nobody has restored from is not a backup. An update nobody applied is an unpatched surface. Uptime monitoring that alerts into a channel nobody reads is decoration.
On a hosted service, the provider handles updates and uptime, and your data passes through their environment. You are trading control for someone else carrying that list. The honest comparison is not “server price versus subscription price” — it is “server price plus maintenance hours versus subscription price”. If nobody on the team has debugged a failed restart at an inconvenient hour, those hours are real, and they tend to land on whoever is most senior.
The trap is treating this as a pure infrastructure argument. It is a staffing argument. Who, by name, is on call when the instance stops responding on a Saturday? If the answer is a person with a full-time job elsewhere, self-hosting is a decision you have made on paper only.
Data residency and PDPA
Self-hosting makes residency concrete: your workflow data stays on infrastructure you choose, in a location you choose. That is a genuine advantage when contracts, customers or internal risk appetite require it. It is also a claim you have to be able to defend. “The data stays with us” is only true if the server, its backups and its snapshots are all actually in the place you say they are.
PDPA keeps the organisation accountable for personal data it handles, including data it passes to a third party. So the compliance question is not “self-hosted or cloud” but “can we describe, in writing, where the data goes and who can reach it?”. On a hosted service, your data passes through the provider’s environment, which makes the provider part of your answer: you need to know their region, their access controls and how long they retain things. On self-hosting, you are the provider — the same questions apply, with nobody to escalate to but yourself.
Scaling limits, and what actually breaks
Scaling arguments are usually overstated for small teams. Workflow volume tends to grow slowly and predictably, and the first real ceiling is rarely throughput — it is the number of automations one person can still understand. That said, the two models fail differently. Self-hosted capacity is yours to plan: when a workflow grows heavier, the upgrade and the downtime window are decisions you schedule. On a hosted service the environment is the provider’s responsibility, so your job narrows to knowing what each workflow does and what data it touches.
What breaks small teams is neither capacity nor uptime. It is the absence of an owner. A workflow nobody documented, running on credentials nobody rotated, on infrastructure nobody monitored, is the same liability whether it is hosted or self-hosted. Settle ownership before you settle hosting.
A decision checklist for a Singapore team
- Is there a named person who can own updates, backups, server security and uptime monitoring — not “the team”, a person?
- Do we need workflow data to stay on our own infrastructure for contractual or residency reasons, and can we prove it stays there, backups included?
- Can we describe in writing where personal data goes and who can access it, for PDPA purposes?
- Is our workflow growth steady enough that a fixed environment stays appropriate through 2026?
- Would our budget be better spent on engineering hours, or on someone else carrying the operational load?
If the first three answers are a clear yes and the fourth is steady, self-hosting is defensible. If any of the first three is “we would have to figure that out”, the hosted route removes a class of failure you are not currently staffed to catch.
For teams that want the hosted model without owning the initial build, the n8n Starter Kit on sakal.shop includes a cloud n8n instance hosted and managed by Sakal Network, 3 custom workflows, initial setup and configuration, 30-day email support, and documentation and handover — S$1,635 including 9% GST (S$1,500 before GST). If your requirement runs wider than n8n, Sakal Network’s AI and document automation services cover workflow builds and document handling. Run the checklist above first, then pick the option your team can actually operate a year from now.
Disclosure: Code n Rob is written by the founder of Sakal Network, which runs sakal.shop.
