n8n vs Make vs Zapier: Tasks, Operations and Executions

Pick n8n if you must self-host or your workflows are step-heavy. Pick Make if you need the most work per dollar. Pick Zapier if you need the widest app catalog and no infrastructure.

The sticker prices mislead: Zapier Professional starts at $19.99/month for 750 tasks, Make Core at $9/month for 10,000 credits, n8n Starter at €20/month billed annually for 2,500 executions (all checked August 2026). Those three units are not comparable.

Key takeaways

• Make stopped billing in operations — credits became the billing unit, converted 1:1, with some AI modules priced dynamically on tokens, file size, pages or run time
• Zapier is no longer a flat 1 task per step: advanced AI models cost 3 tasks, premium models 5, MCP tool calls 2, and Lead Router 5
• n8n bills one execution per workflow run regardless of step count, and sub-workflow runs, manual runs and error workflows do not count at all
• n8n Cloud has no permanent free plan — only a trial. The free option is self-hosting the Community Edition
• n8n is fair-code under the Sustainable Use License, not open source, so hosting client workflows on your own instance needs a separate license

Pick one in a sentence#

One condition decides each choice, and the monthly price is not it.

• Infrastructure and license control decide for n8n. It is the only one of the three that runs on hardware you own, and the only one where the billing unit ignores workflow complexity entirely.
• High step counts at low cost decide for Make. The entry paid plan includes 10,000 credits, and a non-AI module run costs one of them.
• App coverage and zero setup decide for Zapier. The directory lists 9,000+ apps, and every branching, formatting and delay step is free of charge.

A task, an operation and an execution are three different things#

Every comparison that treats these as synonyms produces a wrong answer. Each vendor defines its unit precisely, and the definitions have moved recently.

Zapier: a task is a successful action#

Zapier's help center is explicit: "A task is any successful action that runs in Zapier. Only successful actions count toward your task usage." The unit scales with the number of action steps you draw and the number of times they succeed.

The documented example is a Zap that triggers on a new Gmail email with an attachment, uploads the file to Dropbox and sends a Slack DM. That is two tasks per run, so ten runs use 20 tasks.

Make: an operation is a module run, and credits are what you pay#

Make's definition: "An operation is a single module run to process data or check for new data." Sending five emails through the Gmail module is five operations, not one.

Make formally separated the two terms when credits went live as the billing unit, converting every existing operation balance at a 1:1 ratio. The help center puts the distinction plainly: "Operations count the activities performed in your scenarios, while credits are what you pay for those activities."

For non-AI apps, and for third-party AI apps such as OpenAI, Claude and Gemini, one operation still equals one credit. That equivalence is why the old shorthand mostly still works — and why it quietly breaks on Make's own AI modules.

n8n: an execution is one run of the whole workflow#

n8n's pricing FAQ: "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 — it's still a single execution." The docs repeat it in three words: "An execution is a single run of a workflow."

Only production executions count toward the quota on paid plans, cloud or self-hosted. Those are runs started automatically by triggers, schedules or polling.

So the three units scale with three different variables. Zapier's bill tracks step count, Make's tracks data volume passing through modules, and n8n's tracks how often a workflow fires.

Same job, three bills#

One scenario shape makes the divergence concrete. A scenario watches for new Google Form responses, creates a document for each response, downloads it, and emails it. The trigger returns 10 new bundles in one run.

Applying Make's documented counting rules: the trigger counts as 1 operation, because trigger modules run once per check regardless of how many bundles they return. Each of the three downstream modules runs once per bundle, so 10 operations each. Total: 31 operations for one scenario run.

Applying Zapier's published counting rules to the same job gives a different shape. Ten new form responses fire ten Zap runs; triggers are free; three successful action steps per run cost 3 tasks. Total: 30 tasks.

Applying n8n's rules gives a third answer. A polling trigger counts one execution only when new data is found, and the step count inside the workflow is irrelevant, so a single run that processes all ten responses is one execution.

What never gets billed — the exclusions that decide real cost#

The exclusion lists matter more than the rates.

Zapier bills nothing for: all trigger steps; polling, at any interval; Filter and Paths steps; action steps that error or halt; steps that never run because a filter or path stopped them; the Start a Sub-Zap trigger; and steps for Formatter, Delay, Looping, the Sub-Zap app, Digest, Zapier Manager, Storage, Tables and Forms. Zapier Agents draw on a separate activity quota rather than tasks. A search action set to "Yes (proceed if nothing found)" costs 1 task; set to "No" it costs nothing.

Zapier does bill for: every successful action step, including successful steps inside an error handler path, and previously successful steps that run again when you replay a whole Zap run. Inside a sub-Zap, each action step plus the Call a Sub-Zap and Return From a Sub-Zap steps each cost 1 task.

Make bills nothing extra for trigger volume. The trigger module runs once per check regardless of how many bundles it returns. Everything downstream pays per bundle.

n8n bills nothing for: manual executions; sub-workflow executions, since only the parent top-level execution counts; error workflow executions; polls that return no data; and malformed webhook requests that fail before the workflow starts.

n8n does count one execution for every inbound request that activates a webhook trigger, including requests with an empty body such as {}. A misconfigured sender posting nothing on a loop still burns quota, which is one more reason to validate payloads at the edge — the webhook integration guide covers the pattern.

The multipliers that made 2026 pricing non-flat#

Both hosted platforms added variable rates, and that breaks the flat-rate arithmetic older comparisons rely on.

Zapier's rates page documents per-product task costs, and the task allocation is shared rather than split into per-product budgets: the Free plan's 100 monthly tasks, for instance, are spent across Zap workflows and MCP server actions alike.

Zapier task rates beyond the standard action step (checked August 2026)

Product or step type	Task cost
Standard action step connecting your apps	1 task per successful step
Built-in tools: Formatter, Paths, Filters, Delay, Looping, Sub-Zaps, Digests, Zapier Manager, Storage, Tables, Forms	0 tasks
AI by Zapier, standard models	1 task per AI step; tool calls not available on this tier
AI by Zapier, bring your own key	1 task per AI step, 1 task per tool call
AI by Zapier, advanced models	3 tasks per AI step, 3 tasks per tool call
AI by Zapier, premium models	5 tasks per AI step, 5 tasks per tool call
Zapier MCP tool calls	2 tasks per successful call; failed calls do not count
Zapier Lead Router	5 tasks per successful lead routed
Code by Zapier	1 task per run, plus further tasks when a step runs past the plan runtime allowance on extended runtimes
Zapier SDK actions	Free in beta

Each AI tool call bills separately because the agent picks its tools at runtime, so the cost of one AI step is its model rate plus the model rate times the number of tools it decides to call.

Make's version of the same shift is the split between fixed and dynamic credit usage. Fixed modules charge a set rate — uploading a file with Google Drive always costs 1 credit regardless of file size — while dynamic modules charge on actual usage, most often token consumption, but also file size, page count or run time. Some Make AI Content Extractor modules exceed the default 1 credit per operation because of processing complexity, and the Make Code App costs 2 credits per second of code execution time on paid plans.

n8n adds no per-step multiplier at all. It does meter AI Assistant credits separately: 2,300 per month on Starter and up to 13,700 on Pro. Those credits do not roll over and cannot currently be topped up, and on self-hosted instances you supply your own API key instead.

Free tiers, compared honestly#

Zapier Free is $0/month with 100 tasks a month, unlimited Zap workflows, Tables and Forms, one seat, 15-minute polling, and two-step Zaps only. Multi-step Zaps start at Professional. Free also caps Code by Zapier runtime at 1 second per step and Tables at 2,500 records per account.

Make Free is $0/month with 1,000 credits a month, 3,000+ apps, routers and filters, two active scenarios, a 15-minute minimum interval between runs, a 5-minute maximum scenario execution time and a 5 MB maximum file size. The pricing page states there is no time limit on the Free plan.

Self-hosting and what the license actually allows#

Only n8n self-hosts. Its Business plan is self-hosted at €667/month billed annually for 40,000 executions, Enterprise can be either, and the Community Edition is free to run on your own server.

n8n is not open source, despite near-universal claims otherwise. Its GitHub README states that "n8n is fair-code distributed under the Sustainable Use License and n8n Enterprise License." Fair-code, per faircode.io, means source is openly available and generally free to use, but commercially restricted by its authors.

The operative test in n8n's own docs is narrow and useful: "all use is allowed unless you are selling a product, service, or module in which the value derives entirely or substantially from n8n functionality." Consulting is explicitly permitted, including building workflows and custom features for clients.

The Community Edition also omits real features: Custom Variables, Environments, External secrets, External binary storage, Log streaming, Multi-main mode, Projects, SSO via SAML or LDAP, and workflow or credential sharing. Registering an email address adds folders, debug in editor and custom execution data at no cost.

Business self-hosted subscribers receive a weekly email reporting production execution counts across every instance using their license key. The self-hosting walkthrough covers the deployment side.

Feature matrix#

Zapier, Make and n8n compared on vendor-published terms (checked August 2026)

	Zapier	Make	n8n
Billing unit	Task — any successful action step	Credit — 1 per operation for non-AI apps, dynamic for some AI modules	Execution — one run of the whole workflow
Entry paid plan	Professional from $19.99/mo, 750 tasks	Core from $9/mo, 10,000 credits	Starter €20/mo billed annually, 2,500 executions
Free plan	$0, 100 tasks/mo, two-step Zaps, 1 seat	$0, 1,000 credits/mo, 2 active scenarios, no time limit	None on Cloud — free trial only; Community Edition is free self-hosted
Workflow complexity	Multi-step from Professional up; each successful step bills	Unlimited active scenarios on paid plans; each module run bills	Unlimited steps, and step count never affects the bill
Fastest schedule interval	Free 15 min, Professional 2 min, Team 1 min, Enterprise 1 min	Free 15 min, all paid plans 1 min	Not listed on the pricing page
Users on the entry paid plan	1 seat on Professional, 25 on Team	Teams and team roles begin at the Teams plan, $29/mo	Unlimited users on every plan
Integrations listed	9,000+ apps	3,000+ apps, plus 350+ AI apps	Over 2,000 integrations
Self-hosting	Not offered	Not offered	Community Edition, Business and Enterprise
License	Hosted service	Hosted service	Fair-code, Sustainable Use License, plus Enterprise License for .ee files
Code steps	Code by Zapier: 1 task, extended runtime 1 task per 30s; runtime 1s Free, 30s Pro and Team, 2 min Enterprise	Make Code App: 2 credits per second of execution	Code node included; bash scripts and custom nodes available self-hosted
Throughput controls	Polling interval set by plan	Priority scenario execution on Pro and above; 40 min max run on paid plans	5 concurrent executions on Starter, 20 on Pro, 200+ on Enterprise
Run history retention	Not listed on the pricing page	Full-text execution log search from Pro	7 days on Starter, 30 on Pro, unlimited on Enterprise
Failure handling	Errored and halted steps cost nothing; successful steps in an error path do bill	Store incomplete executions saves failed runs for manual or automatic retry	Error workflow executions do not count toward quota

Verdict for a non-technical solo operator#

Zapier, and the reason is the exclusion list rather than the interface. Filters, Paths, Formatter, Delay and Digest cost zero tasks, so all your branching and reshaping logic is free, and 9,000+ apps means the obscure SaaS tool you depend on is probably already connected.

The cost trap is real and worth naming. Professional is one seat starting at 750 tasks for $19.99/month, so a five-action workflow exhausts the entry tier in 150 runs. Live chat support only begins at the Professional 2,000-task tier.

Verdict for a technical solo builder#

n8n, and the deciding factor is the execution unit combined with the sub-workflow exclusion. You can decompose a large automation into named sub-workflows for readability and still pay for exactly one parent execution.

Starter at €20/month billed annually buys 2,500 executions with unlimited steps and unlimited users. Self-hosting the Community Edition replaces that line item with an infrastructure bill and your own time.

Verdict for a small team#

Split this by workload shape, not headcount.

High-volume, item-heavy data processing goes to Make, because credits track module runs and a single aggregator changes the arithmetic dramatically. Make's own documentation shows a scenario cut from 11 operations to 3 by adding a text aggregator that combined 10 contacts into one email instead of sending 10.

Collaboration and governance across non-technical members goes to Zapier Team, starting at $69/month for 25 users with SAML SSO, shared Zap workflows, shared folders and shared app connections. Engineering-led teams with data-residency or compliance requirements go to n8n Business, self-hosted, which adds SSO via SAML or LDAP, Git version control and separate environments. Companies under 20 employees can check eligibility for n8n's Start-up Plan at 50% off Business.

Migration notes#

No vendor ships a one-click converter between these three platforms. Plan a rebuild, not an import.

1. Inventory by billing unit, not by workflow count

   Count successful action steps per run on Zapier, module runs per bundle on Make, and workflow runs per month on n8n. A 40-Zap account and a 40-scenario account can differ tenfold in cost once you apply the destination platform's counting rules.

2. Move in phases using the webhook bridge

   Make ships Zapier modules specifically for this: a "Receive data (from a Zap)" trigger and a "Send data (to a Zap)" action. Point a Webhooks by Zapier POST action at a Make webhook URL, or catch a hook in Zapier from Make, so a Zap and a scenario hand data to each other during a piecemeal cutover.

3. Rebuild for the destination unit, not the source layout

   Moving to Make, insert aggregators before per-item modules — that is where bundle multiplication is paid or avoided. Moving to n8n, collapse what were separate Zaps into sub-workflows, since only the parent execution counts against quota.

4. Export n8n workflows as JSON or packages

   n8n saves workflows in JSON, with Download, Import from URL and Import from File in the editor. The n8n CLI bundles workflows into a portable .n8np package carrying the folders, projects and references a workflow needs, and checks an import before writing.

5. Strip credentials before sharing any export

   Exported n8n workflow JSON includes credential names and IDs, and HTTP Request nodes may carry authentication headers when imported from cURL. Remove or anonymize that information before the file leaves your machine.

Rebuild error paths at the same time rather than afterward, since each platform prices failure differently — the error handling patterns guide applies across all three.

One honest limitation for each#

Zapier. The entry paid tier is one seat and 750 tasks, and because every successful action step bills separately, cost rises linearly with workflow complexity. That is precisely the workflow shape the other two price better.

Make. Bundle multiplication is invisible on the canvas until the invoice arrives, because a scenario's cost is set by how many bundles the trigger returns, not by how many modules you drew — Make's documented four-module example costs 31 operations per run. The Free plan also caps scenario execution at 5 minutes and files at 5 MB.

n8n. The Sustainable Use License bars white-labeling, hosting-for-clients and collecting end users' credentials, so an agency running client workflows on its own instance needs an Enterprise or Embed license. Self-hosting also means you own upgrades, backups and uptime.

Who this is not for#

If you run one automation that fires a handful of times a month, all three free tiers cover it and this comparison is wasted effort — start with Zapier Free and stop reading.

If your requirement is a specific integration that only one platform lists, availability beats billing math. Check the app directory first and let it decide.

If you need audited data residency or a signed enterprise agreement, none of the published plan tiers answer the question. Those terms live in Enterprise contracts on all three, which are quoted rather than listed, and the tool evaluation checklist is the better starting point.

The recommendation#

For most people reading a three-way comparison, the answer is Make Core at $9/month for 10,000 credits. It is the cheapest path to real volume, its unit is the easiest to predict once you understand bundle multiplication, and the free plan has no time limit while you test.

Choose Zapier instead when app coverage decides the project, or when non-technical colleagues will own the automations — 9,000+ apps and zero-cost logic steps are worth the higher per-step price. Choose n8n when the workflow is step-heavy, when unlimited users matter, or when the data cannot leave your infrastructure.

Do the arithmetic in the destination platform's own unit before you commit. Ten form responses cost 31 operations on Make, 30 tasks on Zapier and one execution on n8n — the same job, three bills, and no exchange rate between them.