Zapier bills you per task. That one design decision is the real reason most people go looking for a Zapier alternative — not missing features, not the interface, but a bill that grows with volume even when the work each automation does is trivial. If your flows are high-volume and simple, Make is usually the cheapest like-for-like migration. If you have someone who can run a container, self-hosted n8n is the cheapest thing that exists at real volume. And if you run twelve fiddly automations a month, you should stay where you are.
The rest of this piece is mostly about counting units, because that is what actually decides your bill.
The thing nobody explains: you are not paying for automations, you are paying for units
Every automation platform charges you for a unit of work. The sticker price on the pricing page is close to meaningless until you know what the unit is and how many of them your workflow burns per run.
Zapier: the unit is a task
On Zapier, a task is roughly "one action that successfully did something." Triggers don't count. Filters that stop a run generally don't count. But every step after the trigger that actually fires — create the record, send the message, update the row — does.
The consequence is linear and unforgiving: a five-action Zap costs about five times what a one-action Zap costs, for the same trigger event. Long, useful workflows are exactly the ones Zapier prices most aggressively. Teams solve this the wrong way, by splitting a good workflow into a worse one to shave actions.
Make: the unit is an operation
Make counts an operation every time a module runs — including the trigger, including each pass through an iterator, including each branch of a router that actually executes. On paper that sounds worse than Zapier, and per-run it often is: the same workflow may burn more operations than tasks.
The reason people still save money is that operations are sold in far larger bundles for comparable money. You use more units, but each unit costs dramatically less. That is the entire arbitrage, and it is why the Make vs Zapier comparison usually resolves in Make's favour for anything running thousands of times a month.
Where it stops being free money: array-heavy work. If you iterate over a 200-row response, you just spent 200 operations. Make's model punishes data-processing loops the way Zapier's punishes long chains.
n8n: the unit is an execution — or nothing at all
n8n counts one execution per workflow run, regardless of how many nodes fire inside it. A 30-node workflow and a 3-node workflow cost the same. That is the cheapest counting unit in the category by a wide margin for anyone building substantial flows.
Self-hosted, there is no per-unit cost at all. You pay for a server and your own time. That "and your own time" is doing a lot of work in that sentence — see below.
The same workflow, modeled three ways
Take a workflow every small team has: a form submission comes in, you enrich it, write it to the CRM, notify a channel, and append a row to a sheet.
- Zapier sees four billable actions per submission. Ten thousand submissions a month is forty thousand tasks. Your bill is a function of your success.
- Make sees the trigger plus five or so module runs — call it six operations per submission, sixty thousand operations. More units, but they are bought in bundles that make the absolute cost lower at that volume.
- n8n sees one execution per submission. Ten thousand executions on cloud, or literally zero marginal cost on your own box.
Nobody needs a fake percentage here. Just look at the shape of the three curves: Zapier's bill scales with steps × runs, Make's scales with modules × runs against much cheaper units, and n8n's scales with runs — or with nothing.
The practical rule: the more steps in your average workflow, the more you are being penalised by per-action billing.
The eight alternatives worth naming
| Tool | Billing unit | Best for | Watch out for |
|---|---|---|---|
| Make | Operation (per module run) | High-volume, multi-step flows; visual builders | Loops and iterators burn operations fast |
| n8n | Execution (or free, self-hosted) | Teams with an engineer and long workflows | You own uptime, upgrades, and secrets |
| Activepieces | Task-ish, open source core | Small teams wanting open source without ops pain | Smaller connector library |
| Pipedream | Credit (per compute) | Developers who want to drop into code mid-flow | Feels like a platform, not a no-code tool |
| Windmill | Compute/worker time | Script-first internal tooling and scheduled jobs | Not a Zapier replacement for non-coders |
| Power Automate | Per user or per flow | Microsoft 365 shops, especially with Dataverse | Licensing is genuinely hard to reason about |
| Workato | Recipe / connection tiers | Enterprise integration with governance needs | Priced for enterprise; overkill for a startup |
| Lindy | Agent runs / credits | Automations that need judgment, not just plumbing | Non-deterministic by design — test the edges |
Swipe the table sideways to see every column →
1. Make — the default migration target
The one most Zapier refugees should look at first. It is a visual builder with a genuine data model: routers, iterators, aggregators, error handlers. You can build things in Make that would take three chained Zaps and a Google Sheet.
Who should use it: anyone running the same workflow thousands of times a month; anyone whose flows have real branching logic.
Who should not: someone who wants an automation live in four minutes without thinking about data shapes. Make asks you to understand your payloads. That is a feature, but it is also a Tuesday afternoon you will not get back.
The honest trade-off: the operation counter is a live wire. An innocent iterator over a large array can quietly become the most expensive thing in your account.
2. n8n — the cheapest at volume, if you have an engineer
Fair-code, self-hostable, node-based, and unapologetically technical. Executions are the billing unit on cloud; on your own infrastructure the marginal cost of another run is a rounding error on your server bill.
Who should use it: teams with at least one person who is comfortable with Docker, backups, and secret management. Anyone whose workflows are long.
Who should not: a solo marketer. "Free" software that you have to keep alive at 2am is not free, and an automation platform that goes down silently is worse than no automation platform.
The honest trade-off: you are trading a subscription for an operational responsibility. That is a good trade for some teams and a terrible one for most.
3. Activepieces — open source without the ops tax
Open-source, MIT-licensed core, with a managed cloud if you don't want to host. It has been aggressive about AI-native pieces and agent-ish steps. The connector library is smaller than Zapier's — that is the whole story with any challenger, and it is the one thing you must check before you commit.
Who should use it: small teams who want open source but don't want to become sysadmins.
Who should not: anyone whose stack depends on a long tail of niche SaaS connectors. Go check the directory before you plan a migration, not after.
4. Pipedream — for people who were going to write code anyway
Pipedream lets you drop a Node or Python step into the middle of a workflow with the connected account's auth already handled. That is the pitch, and it is a good one.
Who should use it: developers who find no-code builders slower than just writing the twelve lines.
Who should not: non-technical operators. The moment something breaks you are reading a stack trace.
5. Windmill — workflows as scripts
Windmill treats scripts as the primitive and builds workflows, schedules, and internal UIs around them. It is closer to an internal-tooling platform than a Zapier competitor.
Who should use it: engineering teams replacing a pile of cron jobs.
Who should not: anyone hoping to connect a form to a CRM this afternoon.
6. Microsoft Power Automate — the answer if you're already Microsoft
If your company lives in Microsoft 365 — SharePoint, Teams, Outlook, Dataverse — Power Automate is often already paid for inside a licence you hold. That changes the maths completely.
Who should use it: IT-governed Microsoft shops.
Who should not: anyone outside that ecosystem. Its connectors to the wider SaaS world are fine; its licensing model is the single hardest thing to reason about in this entire category. Get it in writing from your reseller.
7. Workato — enterprise, and priced like it
Governance, audit, environments, approvals. If those words are on your requirements doc, Workato belongs on your list and most of this article's cost analysis doesn't apply to you, because you're not optimising for price.
Who should not: a startup. You will pay for a compliance apparatus you do not have to satisfy yet.
8. Lindy — when the step needs judgment
Different shape of tool. Lindy is an AI agent platform: instead of "if this, then exactly that," you describe an outcome and let a model handle the messy middle — triaging an inbox, qualifying a lead, drafting the reply. There is a growing overlap between this and classic automation, and it is worth understanding the difference before you choose. The rest of the field is covered in AI agents.
Who should use it: teams whose "automation" keeps failing because the input is unstructured human text.
Who should not: anyone automating a deterministic process. A model that is right most of the time is a liability where a rule would be right every time.
When you should just stay on Zapier
This is the section the alternatives listicles skip because it doesn't sell anything.
Stay if: your total volume is low; your workflows are short; you depend on a connector nobody else has; or the person maintaining automations is not technical and their time is worth more than the subscription. Zapier's app catalogue is still the largest in the category, and "the integration simply exists" beats "the platform is cheaper per unit" every single time.
Migration is not free either. Rebuilding two dozen Zaps takes real days, and the failure modes move — you will find out which of your Zaps were quietly retrying and papering over an upstream bug. Price that in.
A five-minute decision rule
- Export your task usage for the last three months. Not your plan — your usage.
- Count the average number of actions in your busiest five workflows.
- If that number is 1–2 and volume is modest: stay on Zapier.
- If that number is 3+ and volume is high: model it in Make. Same builder metaphor, cheaper units, better data handling.
- If it is 5+, you run thousands of executions, and you have an engineer: n8n, self-hosted, is the end of the conversation.
- If your bottleneck is that the input is unstructured — emails, tickets, form free-text — you don't have a pricing problem. You have an agent problem.
Then check the connector list before you cancel anything. Every migration horror story starts with someone who assumed the integration would be there.
FAQ
Is Make actually cheaper than Zapier?
For high-volume, multi-step workflows, usually yes — not because the plans look cheaper but because the billing unit is cheaper. Zapier charges per action taken; Make charges per module run but sells those runs in far larger allowances. Where Make loses is data-heavy loops, since every iteration consumes an operation. Model your own busiest workflow before switching, and check current pricing on both sites, since these plans change often.
What is the best free Zapier alternative?
Self-hosted n8n or Activepieces, if you can run a container — both have free, open-source cores, and the marginal cost per workflow run is effectively zero. If "free" has to mean "free and hosted for me," you're really comparing free tiers, and those are all deliberately small. A hosted free tier is for testing an idea, not for running a business process.
Can I move my Zaps to another platform automatically?
No. There is no reliable one-click importer between these platforms, whatever a landing page implies. Migration means rebuilding each workflow by hand, re-authorising every connection, and re-testing the error paths. Budget real time for it, and migrate your highest-volume workflows first so the savings start immediately.
Does Zapier count triggers as tasks?
Triggers themselves are generally not billed as tasks; the actions that follow them are. That's why polling triggers on a busy sheet aren't what run up your bill — long action chains are. The practical implication is that shortening workflows saves money and lengthening them costs money, which is a strange incentive to have inside a tool designed to eliminate manual steps.
Is n8n really free?
The self-hosted community edition is free to run under its fair-code licence, but you supply the server, the monitoring, the backups, and the upgrades. If a workflow dies at midnight, nobody is paged except you. n8n Cloud removes that burden and bills per execution, which is still the most favourable counting unit in the category for long workflows.
Which automation tool should a non-technical person use?
Zapier or Make, in that order of gentleness. Both are visual, both are hosted, and neither requires you to think about infrastructure. The open-source options in this list are cheaper precisely because they move work onto you, and if you cannot debug a failed webhook, that saving is illusory.
Where to go next
If you're rebuilding your stack rather than just re-pricing it, start with AI automation tools and no-code AI tools — the two categories where most of this decision actually lives. The full field of Zapier alternatives is worth a scan before you commit, and our guide to the best no-code AI tools covers the builders that pair well with whichever engine you land on. Solo operators paying for all of this out of one bank account should also read the freelancer AI stack. Running something we don't list yet? Submit a tool and we'll take a look.
