Make vs n8n (2026): Which Should Solo Builders Actually Use?
| Tool | Best for | Starting price | Free trial |
|---|---|---|---|
| Make (formerly Integromat) | non-technical automation, connecting CMS/publishing to AI APIs | $9/mo (Core, shown on the pricing page for 10,000 credits a month; we did not check whether that is the monthly or the annual price) | Free plan with no expiry, no credit card required at signup (confirmed firsthand Aug 2026). Re-checked on Make's official pricing page and help center Oct 12 2026: 1,000 credits a month (the unit formerly called operations), at most 2 active scenarios, shortest schedule 15 minutes, 5-minute maximum run, 7-day execution log, 512 MB data transfer, 5 MB maximum file. Our dashboard on Oct 12: 990/1,000 credits left, 0/2 active scenarios. A trigger check costs 1 credit even when nothing is new (measured in our run history Oct 11 2026). |
| n8n | research agent pipelines, publishing automation | Starter 20 EUR/mo billed annually or 24 EUR/mo billed monthly (shown on our expired-trial dashboard, Oct 12 2026; our August "$24/mo" was most likely this monthly price), 2.5k executions a month. Pro 50 EUR/mo billed annually, 10k executions (n8n.io/pricing, Oct 12 2026). | Cloud: 14-day free trial with Pro features, no credit card required (signup page and pricing FAQ, Oct 12 2026), capped at 1,000 executions with a 180-second execution timeout; the Business-plan trial needs a card. One trial per account as far as we could see: our August account showed "Your trial has expired" with the workflow paused and no way to start another trial, only paid plans (data kept about 100 days). Self-hosted Community Edition is free (not tested). |
We didn’t compare these two from spec sheets. We built the identical automation in both — watch this site’s own RSS feed, send an email when a new item appears — and documented what actually happened in each. Full writeups are in our Make review and n8n review; this is the head-to-head.
Getting to a working automation
Make required manually dragging out a trigger module, an action module, and wiring them together — intuitive, but every field (recipient, subject, body) is filled in by hand.
n8n’s AI assistant built the entire two-node workflow from one plain-English sentence. It asked one clarifying question (which email service), then constructed the trigger, the action, and the connection between them itself. For a solo builder without an automation background, this is a real difference in time-to-working- automation, not a marginal one.
What broke, and how visible the failure was
Both tools hit a real problem during setup — this is the part spec-sheet comparisons never capture.
- Make: granting only a minimal “send” OAuth scope for Gmail made the Gmail module fail with
[403] Request had insufficient authentication scopes. (Our August version said the scenario reported “completed” while the email silently never sent; our October 11 re-test didn’t support that. See the correction in our Make review.) - n8n: we granted broader Gmail scope upfront (a lesson carried over from the Make test), and hit no equivalent failure. More notably, before we even asked, n8n’s assistant ran a simulated verification and explicitly said so — “no real email went out during testing” — rather than presenting a simulated pass as a real one.
On this specific dimension, n8n’s tooling was more transparent about the difference between “looks done” and “actually done.”
The polling trap: real in Make, not in n8n (corrected October 12)
(Correction, October 12, 2026: our August version said n8n’s one-minute default would use up a free allowance in about 17 hours, worse than Make. That was wrong. n8n’s documentation says polling triggers such as the RSS Feed Trigger count an execution only when they find new data, and our exported workflow uses that trigger. n8n Cloud also has a 14-day trial rather than an ongoing free plan. We have not measured n8n’s counting on a live account ourselves.)
| Default interval | What an empty poll costs | |
|---|---|---|
| Make (free plan, 1,000 credits/month) | Every 15 minutes | 1 credit each, so about 10 days to use up the allowance (confirmed in our run history, October 11-12) |
| n8n (14-day trial, 1,000 executions) | Every 1 minute | Nothing, per n8n’s documentation; only polls that find a new item count |
n8n’s AI-assistant-generated default was still six times more frequent than what we needed. Adjusting it in n8n was trivial, a plain-English request applied instantly without re-verifying the already-tested workflow. Adjusting it in Make means finding the schedule setting in the trigger module manually, and in Make it matters more, because every check costs a credit. The empty-poll guide has the measured numbers.
Which should you use?
- Pick n8n if you want to describe an automation in plain English and have it built for you, and you check the polling interval it picks — or you want the option to self-host the Community Edition for free indefinitely once you outgrow the trial.
- Pick Make if you’d rather see and control every module yourself from the start, and you’re comfortable digging into isolated module tests when a step doesn’t run.
For our own pipeline, both are staying in the stack for now — n8n for building new automations quickly from a plain description, Make for scenarios where we want to see the full wiring while we’re still learning what we actually need.
Two things worth checking before you choose: what happens when n8n’s trial ends (what you see when the n8n trial expires), and how far Make’s free credits go on your schedule (what Make’s free plan actually gets you and the Make credit calculator).