For about the cost of a mid-range phone plan, an Australian SME can run a couple of thousand automated jobs a month on n8n's hosted platform — invoice routing, CRM syncing, quote follow-ups, the overnight reconciliation that someone currently does by hand on Monday morning. n8n's published pricing starts at €20 a month billed annually for 2,500 executions, and the self-hosted Community edition licence is free. Both of those numbers are true, and neither of them is the number that decides the question.
The licence is almost always the smallest line in an automation budget. What actually determines whether n8n is cheap or expensive for your business is how many times your workflows fire, and who is accountable when the instance stops firing at eleven o'clock on a Tuesday night.
The unit you are buying is the execution
n8n prices its cloud plans by monthly workflow executions rather than by steps, nodes or seats. The visible tiers are €20 a month billed annually for 2,500 executions, €50 a month for 10,000, and €667 a month for 40,000, and all plans include unlimited workflows, steps and users.
That last part matters more than it sounds. A workflow with forty nodes — an HTTP call, a router, three branches, retries, a Slack notification and a database write — counts the same as a workflow with two nodes. This is structurally different from tools that bill per operation or per task, where a single business process can consume a dozen billable units every time it runs. If you have priced automation on Make's plan structure or something similar, the mental model does not transfer, and Make adjusted its own extra-credit costs and plan limits effective 6 November 2025, so older comparison articles are stale on both sides.
The complication is that n8n's pricing changed too. In August 2025, n8n removed active-workflow limits and moved to unlimited users, workflows and steps, introduced a self-hosted Business plan, and shifted Enterprise pricing toward executions rather than workflow count. Any cost comparison written before that date is describing a product that no longer exists.
Model your executions before you pick a tier
The number that breaks budgets is not the number of automations. It is the number of times they run, and the usual surprise is polling.
A workflow triggered by a webhook fires when something happens. A workflow on a schedule fires whether or not anything happened. Set a trigger to check an inbox, a folder or an API endpoint every five minutes and it runs 288 times a day — roughly 8,640 times a month, from one workflow, before it has done anything useful. Drop that to every minute and you are at about 43,200. Three or four polling workflows at that cadence will consume an entry tier before a single customer-facing process has run.
The fixes are mostly design decisions rather than plan upgrades:
- Use webhooks wherever the source system offers them. Most modern SaaS does.
- Poll on business hours, not around the clock, where the data only changes during business hours.
- Batch. One scheduled run that pulls fifty new records is one execution; fifty triggered runs are fifty.
- Watch what sub-workflows and error workflows do to your counts, and confirm against your own execution list rather than assuming.
The practical approach is to run the thing for a fortnight before committing to a tier. n8n offers a free cloud trial, and the execution list will tell you your real volume far more reliably than an estimate built in a spreadsheet. Designing triggers so the execution count stays flat as volume grows is a large part of what makes an n8n build economical to run rather than something that quietly escalates a tier every quarter.
The free licence sits next to several paid ones
Self-hosted n8n is not one thing. The documentation distinguishes Community, Registered Community, Business and Enterprise editions alongside Cloud. Community is the free self-hosted edition. The paid self-hosted tiers add capabilities including SSO, environments, projects, version control and scaling features, and Enterprise is positioned as the option for advanced security, scaling and governance.
This catches people out. The assumption is "self-hosted equals free equals feature-complete", and it is not. If your requirement includes single sign-on, separate development and production environments, Git-backed version control of workflows, or granular project permissions across a team, you are looking at a paid self-hosted plan, not the Community edition. Worth checking the Community edition feature list against your actual requirements before assuming the licence line is zero.
What self-hosting costs to run, line by line
Here is the shape of the comparison once you put the licence in its proper place.
| Cost line | n8n Cloud | Self-hosted |
|---|---|---|
| Licence | €20 to €667+ per month by execution tier | Free on Community; paid on Business and Enterprise |
| Compute and database | Included | VPS or container hosting plus a managed Postgres instance |
| Scaling under load | Included | Queue mode, Redis and additional worker processes |
| Backups and restore testing | Included | Yours to configure, run and actually test |
| Version upgrades | Applied for you | Yours, on a release cadence that moves quickly |
| Security patching and credential key management | Included | Yours |
| Monitoring and alerting on a stalled instance | Included | Yours to build and route to someone |
| Someone accountable at 11pm Tuesday | The vendor | A named person in your business |
None of those self-hosted lines are exotic. They are all normal infrastructure work. The point is that they exist, they recur, and they land on someone's plate permanently. The costs most often underestimated in an automation program are not the platform fee at all — they are exception handling, duplicate prevention, reconciliation, alerting, credential rotation, API rate limits and human approvals, plus ownership of the whole thing after whoever built it moves on.
Where the break-even actually sits
On pure arithmetic, cloud wins comprehensively at the low end. At 2,500 executions a month, €20 billed annually does not buy you an hour of anyone's attention, let alone a server, backups and a monitoring stack.
The arithmetic tilts as volume climbs. A business running well past 40,000 executions a month is looking at a four-figure monthly platform bill, which is the point where a self-hosted instance on modest infrastructure starts to look cheaper on paper. If your workloads are heavy — high-frequency polling, large batch jobs, AI steps that fan out into many tool calls — you reach that point sooner than you expect.
But the crossover on the spreadsheet is not the decision. Three questions decide it:
The Australian currency and tax footnote
n8n publishes in euros rather than Australian dollars. The landed cost depends on the exchange rate on the day, whatever conversion fee your card issuer adds, and how GST is treated on your particular arrangement — so €20 a month is meaningfully more than €20 a month by the time it hits the ledger, and the gap widens on the larger tiers. If you are building a business case at the €667 level, get the actual AUD charge from a real invoice rather than a converter, and confirm the GST position with your accountant. The same applies to the separate bill for any LLM provider your workflows call, which is priced independently of the automation platform and scales with tokens rather than executions.
The first move
Before you compare anything, count. Open the execution list on whatever you are running now — a trial instance, a competitor's tool, a Power Automate tenant — and pull thirty days of history. Sort by workflow. You will almost certainly find that two or three polling triggers account for most of the volume and almost none of the value, and that fixing those changes which tier you need.
If you would rather have the number than derive it, our automation ROI calculator will model your execution volume, platform tier and run-cost against the labour the workflows replace, and we will give you a scoped build estimate and a plain answer on whether cloud or an Australian-hosted instance is the right call for your situation — including the answer "your volumes do not justify this yet", when that is the honest one. Typical builds land in two to four weeks, on cloud or on infrastructure you control.