3 Automation Platforms, 3 Meters: n8n vs Zapier vs Make in 2026

Every automation comparison turns into a feature table, and the feature tables all look the same: a canvas, a library of integrations, an AI step you can drop in anywhere. What actually differs is the unit on the invoice. Zapier charges per action, Make charges per credit, n8n charges per completed run. Run the same ten-step workflow a thousand times and those three meters produce three completely different bills.

That is the whole comparison. Everything below is the detail behind one sentence: the meter decides your cost shape, not the feature list.

1. Zapier: 9,000 Apps and a Meter That Counts Every Action

Zapier is the catalogue play. It connects more apps than anyone else, it needs no code, and it has been the default answer to "connect these two things" for a decade. Its pricing unit is the task, and a task is awarded to each successful action a Zap performs.

The accounting is more generous than it first looks. Triggers are free - the event that starts a Zap never counts. Polling for new data is free, so a Zap that checks Dropbox every five minutes and finds nothing costs nothing. A whole list of built-in steps also runs without charging: Filter, Formatter, Paths, Delay, Looping, Sub-Zap, Digest, Storage, plus Zapier's own Tables and Forms. Only the actions that actually move data between your apps bill.

Pricing: Free is $0 with 100 tasks per month, two-step Zaps only, one user, and a 15-minute polling interval. Professional starts from $19.99 per month and unlocks multi-step Zaps, webhooks, AI fields, unlimited premium apps and two-minute polling. Team starts from $69 per month for up to 25 users with one-minute polling and SAML SSO. Enterprise is custom. Annual billing saves up to 33%. If a Zap exceeds your task tier, you are switched to pay-per-task billing at a higher per-task rate; if that switch is turned off, the Zaps pause until the next usage period.

Where it falls short: The task is not always one per action, and Zapier says so in its own documentation - AI model tier, code runtime and connector type each change how many tasks a step consumes, and an MCP tool call costs two. So the headline "one action, one task" is a floor, not a rate. Second, 100 tasks is roughly three actions a day, which is a demo rather than a plan. Third, Free and Professional are single-user plans; collaboration starts at Team, which is where the price jumps from $19.99 to $69.

2. Make: A Credit Meter, a Visual Canvas and a 15-Minute Floor

Make approaches the same job from the other direction: instead of a linear list of steps, you get a scenario canvas where branches, routers and iterators are visible as geometry. It covers 3,000+ standard apps and ships AI agents and an MCP server on every tier, including Free.

Its meter is credits, and that unit is recent enough to still confuse people. Make's own pricing FAQ states that as of August 27, credits are the billing unit, and that most module actions consume one credit while advanced features built on Make's AI Provider can consume more. Two exceptions are worth knowing because they are where heavy scenarios hide their cost: router modules and error-handler modules (Rollback, Break, Resume, Commit, Ignore) consume nothing, but the Make Code App charges 2 credits per second of code execution time. A ten-second JavaScript step is 20 credits before it has touched a single app.

Pricing: Free is $0 with 1,000 credits per month, two active scenarios, a 15-minute minimum interval between runs, a five-minute cap on execution time and 512 MB of data transfer. Core is $9 per month, Pro $16 and Teams $29 - each quoted at the 10,000 credits per month level, so the entry price moves with the credit tier you choose. Annual billing saves 15% or more. Running out of credits does not silently stop billing you more; scenarios stop, and you either buy extra credits in 1,000 or 10,000 bundles or enable auto-purchase, which is available on Core, Pro and Teams.

Where it falls short: The 15-minute floor on Free is the detail that decides usability. Anything that needs to react within a minute - a form submission triggering a reply, a webhook triggering a Slack alert - is not viable there regardless of how many credits remain, and Free also caps you at two active scenarios and a five-minute run. The credit model is fair for branching logic but expensive for long linear scenarios: a twenty-module scenario burns twenty credits per run, where n8n would charge one for the same run. And because credits are consumed per module rather than per completed run, failed mid-scenario executions still cost.

3. n8n: One Charge Per Run, Unlimited Steps, or Your Own Server

n8n meters the opposite way: once per workflow execution, with unlimited steps inside it. Its pricing page makes the contrast explicit - unlike tools that charge per step or per user, n8n lets you build freely and pays only when a workflow runs from start to finish. Users and workflows are unlimited on every plan, and every integration is included, so there is no "premium apps" tier to upgrade into.

Pricing: The Community Edition is self-hosted, available on GitHub, and carries no licence fee - you pay for the machine and for the model API keys your AI nodes call. Cloud Starter is 20 EUR per month billed annually with 2,500 workflow executions, one shared project, five concurrent executions and 1,600 assistant credits per month. Pro is 50 EUR per month with 10,000 executions and up to 50 concurrent executions. Business is 667 EUR per month with 40,000 executions and is self-hosted only, despite being a paid tier. Enterprise is custom with unlimited projects and 200+ concurrent executions. Annual billing saves 17%, and overage fees are capped rather than open-ended.

Where it falls short: Self-hosting is the entire value proposition for a lot of buyers, and it is also the entire cost. You run the database, the queue, the upgrades and the backups, and you are the person paged when a workflow stops firing at 2am - the software being free does not make the on-call rota free. On the hosted side, prices are quoted in euros, so a dollar-based budget carries an exchange-rate risk that Zapier and Make do not impose. And the Business tier being self-hosted only is a real constraint: the tier with the best cost per execution is the one where you still own the infrastructure.

How to Choose

Do the arithmetic on your own workflow before you compare plan names. Take a ten-step automation that runs 1,000 times a month:

• Zapier - roughly 1,000 x 10 = 10,000 tasks, before any step that costs more than one task.
• Make - roughly 1,000 x 10 = 10,000 credits, minus routers and error handlers, plus anything the Code App or AI Provider adds.
• n8n - 1,000 executions on the hosted plans, or nothing beyond your own server costs on Community Edition.

That one calculation explains why the platforms feel so differently priced in reviews: a two-step automation is cheap everywhere, and a long automation is cheap only where steps are free. If your workflows are short and numerous, Zapier's per-action meter is the one that matches; if they branch a lot and you want the canvas, Make's credit meter stays reasonable; if they are long, or you would rather own the server than rent it, n8n is the meter that stops counting steps.

The trap to avoid is comparing free tiers as if they were the product. Zapier's 100 tasks, Make's 1,000 credits behind a 15-minute floor, and a self-hosted n8n instance are three different things wearing the word "free": a demo, a small workload, and a server you now maintain. Pick the meter you can live with at volume, then check what the free tier is actually shaped like.

One more consistency note, because it trips people up when they read all three pricing pages in one sitting: Make's pricing page still uses the word "operations" in its usage-allowance section, while credits are the billing unit everywhere else. The two refer to the same thing in different generations of the page - do not treat them as two separate meters.