Three automation platforms, one six-step flow, and three different things being counted. Every figure below was read from the vendor's own pricing page and logged in our catalogue on 19 November 2026.
A pricing page answers the wrong question. It tells you what a month costs, and the money that actually leaves the account depends on a unit the page barely mentions: a task, a credit or an execution. Zapier counts one, Make counts another, and n8n counts a third, so a workflow that looks like the same purchase on all three is not the same purchase at all.
I maintain the pricing catalogue at AI-SaaSRadar, operated by MMS Consulting Group (엠엠에스컨설팅), an independent software directory. For this piece I read the three vendors' plans and their internal limits, then put one identical flow through each of their meters. No estimate enters the arithmetic; the divisions and multiplications are printed.
The meter decides the bill, not the plan name
Zapier's meter is the task. Every action inside a Zap counts as one, so a six-step Zap spends six tasks per run. That single sentence explains most of the complaints about the product: the plan is sold in thousands of tasks, and a workflow with a filter, a formatter, a lookup and two writes eats them fast.
Make's meter is the credit, and the company unified on it on 27 August 2026. One module action costs one credit, but the modules do not all cost the same: some that call an AI provider bill more than one, and an iterator multiplies by the number of rows it walks. A six-module scenario is six credits per run, until someone adds a loop over 200 spreadsheet rows.
n8n's meter is the execution, and the whole workflow is one. A 40-step workflow and a two-step workflow both count once. The limits that actually stop you are elsewhere: concurrency, saved executions, log retention and the per-run time cap.
Three platforms selling "automation" are selling three different units of work. The price list cannot tell you which one is cheaper for your job, because it never says how many units your job is.
What the free plans are worth in finished work
Now price the same thing on each: a six-step flow, run once from trigger to last write.
Free allowances converted into completed runs of a six-step flow. Read from ai-saasradar.com, 19 November 2026.
Zapier's free plan allows 100 tasks a month, and the same plan is limited to two-step Zaps. The second limit matters more than the first: a six-step flow cannot be built on it at all, so the free allowance of 100 tasks converts into zero finished runs of this job. Multi-step Zaps, premium apps and webhooks are paid features, which is why the entry to the product is Professional at $19.99 a month on yearly billing. If you divide 100 by six anyway, you get sixteen runs, or roughly three actions a day. It is a trial of the interface, not a working automation.
Make's free plan stops at a monthly credit allowance that our catalogue records as around 1,000 credits. At six credits a run, that is about 166 completed runs of the same flow — around five a day, before any iterator section multiplies the count. Paid entry is Core, $9 a month.
n8n sells no free cloud tier. There is a 14-day trial with 1,000 executions, five concurrent executions and a 180-second cap per run, and after those 14 days the cloud is paid from Starter at €20 a month billed yearly. The self-hosted Community Edition is the only thing that stays free, and it stays free forever — but you own the server, the upgrades and the backups.
So the same free plan is worth zero, 166 or 1,000 runs of your job. The number that matters is not on any pricing page, and it is not comparable until you fix the shape of the workflow.
What the category says about the price
Entry price still tells you something, as long as you know where it sits.
Entry price of the 22 tools in the Automation category, cheapest first. Read from ai-saasradar.com, 19 November 2026.
Across the 22 automation tools we track, the median published entry price is $16.50 a month. The range runs from $5 (Groq LPU) to $199 (Relevance AI), and the three platforms in this piece are unremarkable in it: Make is fourth cheapest at $9, Zapier is thirteenth at $19.99, n8n is fifteenth at $20. Six of the 22 start at $10 a month or less, and anyone paying $20 for one seat is buying something other than the lowest price.
That something is usually coverage. Zapier's catalogue of connected apps is the reason teams pay 2.2 times Make's entry price, and the reason an automation survives when the app you need is obscure. Make's routers and iterators let you draw branching and looping that Zapier's linear Zaps cannot express. n8n gives you the same logic with code nodes and a licence to run it yourself.
A seat multiplies whatever the meter already was
The second multiplier is the one budget owners forget, because it looks linear and therefore harmless until a team grows.
Entry plan price times seats. Tax and usage overage are not included. Read from ai-saasradar.com, 19 November 2026.
One person costs $9 on Make, $19.99 on Zapier and $20 on n8n. Five seats cost $45, $99.95 and $100. Twenty seats cost $180, $399.80 and $400.
That last row is where the gap stops being a rounding error and starts being a line item someone has to defend. It is also where the units diverge again: on Zapier and Make the seat is charged in addition to the metered allowance, so a five-person team sharing one Professional plan still runs out of tasks and credits. On n8n, adding people to a self-hosted instance costs nothing at all, and adding them to the cloud plan costs a plan upgrade.
The costs the page does not print
Each of the three has a wall that appears only after the plan is signed.
Zapier. When the task allowance runs out, the plan switches to pay-per-task billing rather than stopping. Premium apps and webhooks are paid features, so a workflow that depends on them cannot even be assembled on the free plan.
Make. Credits multiply quietly. A six-module scenario bills six credits per run, an iterator over 200 rows bills per row, and a module that calls an outside AI provider can cost more than one credit. An allowance that looked comfortable at design time can disappear in an afternoon of backfill.
n8n. The Starter plan is reasonable until a team needs single sign-on, Git-based version control or separate environments — those arrive on Business at €667 a month. That is a 33-fold step from Starter, and it is the step most self-hosting teams eventually face.
Moving is rebuilding, not exporting
If the meter turns out to be wrong for your work, the exit has a cost too, and it is measured in hours rather than dollars.
A Zap cannot be exported as a file that another tool will load. Make's scenario blueprints export as JSON, but only into another Make organisation. n8n workflows export as JSON and re-import into n8n, yet credentials are encrypted separately and must be entered again, and webhook URLs differ per instance, so every outside service that calls them needs the new address.
The practical version: document the step list, re-authorise every connected account on the new platform, and check that trigger polling intervals and run-history retention behave the way your incident response assumes. Budget a day for a six-step flow, and more if it branches.
How to choose without a spreadsheet nobody updates
Three questions settle most of these decisions, and none of them is "which is cheapest".
Who operates it? If the answer is nobody, self-hosted n8n is the wrong choice regardless of price. The server, the upgrades and the backups become your team's work, and that work has a cost the invoice never shows.
How does the flow grow? If it stays short, Zapier's task meter is cheap and its app coverage is worth the premium. If it keeps adding steps and branches, n8n's execution meter stays flat while Zapier's grows with every step. If it needs looping over rows, Make's visual iterators are easier to build and easier to overspend.
Do the logs have to stay inside your infrastructure? If yes, only the self-hosted option satisfies it, and the rest of the comparison is decoration.
Then run one real job end to end on the free tier, count the units it consumed, and multiply by the number of times you expect to run it next month. That figure, not the sticker, is the price.
The full head-to-head pages, with each vendor's free-tier stopping point written out and dated, are here: Zapier vs Make, Zapier vs n8n and Make vs n8n. The three tool records, with every tier side by side and the verification date for each price, are at Zapier, Make and n8n; the alternatives list at n8n alternatives prices 21 replacements on the same basis, and workflow pricing adds up real stacks step by step.
Prices and limits move often. Where our catalogue and a vendor page disagree, the vendor page is newer; the check date beside each figure says when it was last read.



Top comments (0)