If you are comparing Zapier with Make and n8n, most comparison pages will show you pricing screenshots that are already out of date. The more durable question is how each platform counts usage, because that decides what a given workflow will cost you at scale. What follows is based on the billing rules and licenses as published in September 2026.
The short answer
Zapier is the fastest way for non-technical teams to connect SaaS apps. It has the most integrations and needs the least setup. At volume, it's also the most expensive per step.
Make is a visual builder that handles branching and iteration well, with good error handling on top. It can do more than Zapier in complex scenarios and costs less per step, although the learning curve is steeper.
n8n is the developer-friendly one, and you can drop into code when you need it. It bills per workflow run. It is also the only one of the three you can self-host, which makes it the best fit for long workflows at high volume, run by a team with engineering support behind it.
How each one bills
This is the part most comparisons get wrong.
| Zapier | Make | n8n | |
|---|---|---|---|
| Billing unit | Task: each successful action step | Credit: each module that runs (most cost 1; some AI modules cost more) | Execution: one full workflow run, any number of steps |
| Free steps | Trigger, Filter, Paths, Formatter, Delay, Looping, Storage | Routers, error handler directives, inline functions | Everything inside the run |
| Self-hosting | No | No (Enterprise on-prem agent only) | Yes (Community Edition) |
| Free tier (cloud) | 100 tasks/month | 1,000 credits/month | Self-hosted Community Edition |
Plans and prices change often, so check the current pages before deciding. The units above are what you plug your numbers into. Still, the tool is usually the smallest part of the total bill. Our guide to what AI automation really costs covers the build and upkeep.
A worked example
Take a realistic workflow with 8 steps. A webhook trigger starts it, then a filter and a formatting step run, and 5 app actions follow (look up customer, create record, update CRM, post to Slack, send email). It runs 10,000 times a month, and 20% of runs stop at the filter.
| How it counts | Monthly usage | |
|---|---|---|
| Zapier | 5 tasks × 8,000 runs that pass the filter; trigger, filter and formatter are free | about 40,000 tasks |
| Make | Roughly 1 credit for the trigger on every run, plus 5 per passing run; filters and inline functions are free | about 50,000 credits |
| n8n | 1 execution per run, filtered or not | 10,000 executions |
Now double the workflow to 16 steps. Zapier and Make usage roughly doubles, while n8n stays at 10,000 executions. That is the structural reason n8n wins on cost for long workflows. It's also why Zapier is fine for short ones.
Self-hosting n8n: free license, real costs
Running n8n yourself removes the per-execution bill, and it keeps data inside your infrastructure. It also makes you the operator, and that job has costs you should budget for.
Patching comes first. In January 2026, n8n published a critical advisory (CVE-2026-21858, CVSS 10.0) for unauthenticated file access through webhook request handling, affecting versions 1.65.0 up to 1.121.0. Anyone self-hosting an unpatched, internet-facing instance was exposed. You need a routine for tracking and applying releases.
Scaling is next. High volume needs queue mode with separate workers and Redis, plus a proper Postgres database in place of the default SQLite.
Then there's the everyday operations work of backups and monitoring, along with access control. Workflows hold credentials to every system they touch, so treat the instance like production infrastructure.
For many small teams, n8n's cloud plans are the better deal until volume or data-residency needs justify self-hosting.
Licensing: "fair-code" is not open source
n8n's Community Edition is free to use. Its license is the Sustainable Use License, though, and that is not an OSI-approved open-source license.
In practice, you can use it for your own internal business processes. Consultants and agencies can also build and run workflows for clients on their own instance and charge for that work. What you cannot do without a commercial agreement is let external users build or configure workflows through your product, for example by embedding n8n's editor in a SaaS you sell.
So if your plan is to put automation inside a product your customers configure, talk to n8n about a commercial license or pick a platform built for embedding. Zapier, for example, offers an embed program for this case.
Which one should you pick?
| If your situation is... | Pick |
|---|---|
| Non-technical team, a few simple workflows, many different SaaS apps | Zapier |
| Ops team building complex branching scenarios, cost-conscious | Make |
| Engineering support available, long workflows, high volume | n8n (cloud) |
| Strict data residency, or volume where per-run pricing dominates | n8n (self-hosted) |
| Mission-critical, high-volume integration with strict SLAs | Consider custom code |
That last row matters. Low-code tools are excellent for internal workflows. A core integration, such as order sync between your store and ERP or payment reconciliation, often deserves real code with retries and idempotency, plus tests and monitoring. We often see companies run the long tail of workflows in a low-code tool and move the two or three business-critical ones into code.
Migrating between them
There is no reliable automatic converter between the three. Migration means rebuilding each workflow, which also makes it a good moment to consolidate duplicates and remove workflows nobody uses. Start with the ones that cost the most under your current billing unit, as calculated above.
If you are weighing this for a specific process, like accounts payable, see our walkthrough of automating invoice processing with AI. For recurring client reports built from files that arrive by email and SFTP, see how to automate monthly reports. If new leads sit unanswered while they bounce between inboxes, our scenario on how to automate lead follow up shows where these tools fit and where a person should take over.