Zapier's Professional plan costs $19.99 a month billed annually and covers 750 tasks, where a task is one successful action inside an automation. n8n's Starter plan costs 20 euros a month billed annually and covers 2,500 executions, where an execution is one whole run of a workflow no matter how many steps it contains. Those two allowances sound comparable and are not. Which of the three tools is cheapest depends almost entirely on how many steps each of your automations has, since Zapier, Make and n8n each count a different thing.
The three tools bill for different units
n8n's pricing page states the simplest rule: "An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes." Zapier sits at the other end. Its help center defines a task as "any successful action that runs in Zapier," and only actions that succeed count toward the total. Make charges in credits, and Make's pricing page says credits are its billing unit, that each action a scenario performs consumes a certain number of credits, and that most actions cost 1 credit while some advanced features that call Make's AI provider cost more.
The exemptions matter as much as the definitions. On Zapier, trigger steps, filters and paths are free, as are steps that run in Formatter, Delay, Looping, Digest, Storage, Zapier Manager and the Sub-Zap app, and so are action steps that error or halt. A Zap that triggers on a new form entry, filters out junk entries, reformats a phone number and then writes to two apps therefore costs two tasks per run, not five.
Make counts more finely. Its documentation defines an operation as "a single module run to process data or check for new data," and a module's count depends on how many bundles of data it processes, so a module that handles five rows counts five operations. Trigger modules are the exception and "only run once to check for or retrieve data, regardless of the number of bundles returned." Loops and list processing are where Make bills climb.
What the entry plans cost as of October 2026
Zapier's free plan includes "100 tasks/mo; unlimited Zap workflows/Tables/Forms; two-step Zap workflows" for one seat. Professional starts at $19.99 a month billed annually, or $29.99 billed monthly, for 750 tasks, and at the 2,000 task tier the same plan is $49.00 annually or $73.50 monthly. Team begins at 2,000 tasks and costs $69.00 a month billed annually, or $103.50 monthly, and covers 25 seats. The page notes that "Annual subscriptions are billed yearly at 33% off the monthly rate," and that the 750 and 1,500 task tiers "exist only on Professional; Team begins at 2,000 tasks/mo."
n8n prices its cloud plans in euros. Starter is 20 euros a month billed annually for "2.5K workflow executions with unlimited steps," and Pro is 50 euros a month billed annually for "10K workflow executions with unlimited steps." A cloud trial gives Pro level features and 1,000 executions with no credit card.
Make's free plan gives 1,000 credits a month, which the pricing page says "equals 1 MB of data storage and 1 data store." Paid tiers are Core, Pro, Teams and Enterprise, and the allowance scales with the number of credits you buy rather than sitting at one fixed figure, so read the current number off the pricing page before you budget. Per 10,000 credits a month, Make lists 5 GB of data transfer and 10 MB of data storage.
Price your own workflow before comparing plans
The only honest comparison is one of your own automations, counted three ways. Take a workflow you actually plan to run, estimate how many times it will fire per month, then count how many billable units each tool would charge for a single run.
Example: an intake workflow fires when someone submits a contact form, filters out blanks, reformats the phone number, creates a CRM record, sends a confirmation email and posts a message to a team chat. Say it runs 400 times a month. Zapier: the trigger, the filter and the Formatter step are free, so three billable actions per run. 3 x 400 = 1,200 tasks a month, which does not fit the 750 task tier. Make: six module runs per scenario run. 6 x 400 = 2,400 credits a month, so the 1,000 credit free plan is out. n8n: one execution per run. 400 executions a month, comfortably inside a 2,500 execution plan.
The pattern holds beyond that example. Short workflows that fire often suit Zapier, since most of the scaffolding steps are free there. Long workflows, loops over lists and anything that handles many rows per run point to n8n, where step count does not touch the bill. Make sits between the two and rises fastest on iteration over records, which is the part people tend to underestimate.
Self-hosting n8n shifts the cost onto a server you run
n8n publishes a standard self-hosted version on GitHub under its Sustainable Use License. The license grants a royalty-free, non-transferable right to copy, distribute and prepare derivative works, but only for a company's own internal purposes or for non-commercial and personal purposes. Passing the software to others is allowed "only if you do so free of charge for non-commercial purposes," and files with ".ee." in the name fall under a separate n8n Enterprise License. Running automations for your own company sits inside those terms. Reselling n8n as a hosted product does not.
The bill becomes server hosting and maintenance time. DigitalOcean's cheapest basic droplet is $4.00 a month for 1 vCPU, 512 MiB of memory, 10 GiB of SSD and 500 GiB of transfer as of October 2026. n8n's DigitalOcean guide says a basic shared CPU plan is enough at most levels of traffic, and asks for "at least 4 GB RAM / 2 vCPU" if you add the Assistant sandbox, which runs Docker inside Docker and needs more headroom than n8n alone.
What the server price does not cover is updates, backups, monitoring and the time it takes to restart a container that stopped overnight. Self-hosting is the cheapest option in dollars and the most expensive in attention, so it fits only when someone at the firm is comfortable with a terminal. It does solve one real problem: on a self-hosted instance, client records and run history stay on infrastructure you control, which matters if your automations touch health, financial or client data covered by a contract.
The multipliers that break a budget estimate
AI steps do not cost one unit. In Zapier MCP, "each successful tool call counts as 2 tasks," and failed calls do not count. Make states that some advanced features calling its AI provider consume more than one credit per action. Any estimate built on one unit per step will undercount a workflow that calls a model.
Hitting the ceiling mid month has its own behaviour. Without pay-per-task billing, Zapier says that "When your plan reaches its task limit Zapier will hold any new Zap runs until your billing cycle resets," with email alerts at 80% and 100% of the limit. With pay-per-task billing switched on, extra tasks are charged separately, and tasks are held until the end of the cycle once volume reaches 3x the plan's task limit. A held run means a confirmation email that never sends, so the failure is invisible from the client's side.
What to do this week
- Write down the single workflow you would automate first and estimate its monthly run count from real figures, such as last month's form submissions or invoices.
- Count the steps, marking which ones are triggers, filters or formatting, since those are free on Zapier and billable on Make.
- Multiply runs by billable units for each tool and compare against the tiers above rather than against headline prices.
- Build it on a free tier first and watch the counter for a week. Real volume tends to run higher than the estimate, thanks to retries and test runs.
Two limits on all of the above. Prices on all three pages change, and n8n quotes euros while Zapier quotes dollars, so check both pages on the day you decide and convert the euro figure yourself. None of this is urgent for most very small firms either: Census Bureau Business Trends and Outlook Survey data covering December 14, 2025 to May 3, 2026 found that less than 20% of firms with four or fewer employees reported AI adoption, against 37% of firms with at least 250 employees.
