Make vs n8n vs Activepieces (2026): Same Task, Compared
| Tool | Best for | Starting price | Free trial |
|---|---|---|---|
| Make (formerly Integromat) | non-technical automation, connecting CMS/publishing to AI APIs | $9/mo | Free tier, no credit card required, 1,000 ops/mo (confirmed firsthand Aug 2026; re-checked on the official pricing page Sep 27 2026, unchanged) |
| n8n | research agent pipelines, publishing automation | Starter 20 EUR/mo billed annually (shown in EUR on the pricing page Sep 27 2026, not USD as our Aug figure assumed - may be regional; needsVerification) | Cloud: free trial, no credit card required (confirmed firsthand Aug 2026, trial was 14 days then). Re-checked the official pricing page Sep 27 2026: still says no credit card, but the 14-day length is no longer stated on the page - needsVerification. Self-hosted Community Edition is free forever, no execution/workflow limits. |
| Activepieces | a third free automation option alongside Make and n8n, testing/building without burning a free-tier allowance | $16/mo (Plus, billed yearly) | Free forever, no credit card required (confirmed firsthand Sep 2026, official pricing page and FAQ). 100 credits/day, unlimited flows. Credits are only consumed by live/published flow runs - building and testing steps in the editor consumed 0 credits in our test. Sep 26: the in-app credit box read 1,000 credits resetting Oct 7 (not re-checked against the pricing page), and one AI Chat run took it to 0 credits / 100% used. |
This isn’t a feature-list comparison. We built the exact same real automation - watch this site’s own RSS feed, send a Gmail notification when a new item appears - in all three tools, on fresh free-tier accounts, and logged what actually happened in each. Full writeups: Make, n8n, Activepieces.
The same task, side by side
| Make | n8n | Activepieces | |
|---|---|---|---|
| How it was built | Manual drag-and-drop, wired by hand | AI assistant, built from one sentence | Manual drag-and-drop |
| Gmail OAuth scope requested | Minimal (“send” only, as configured) | Broad (view, compose, send), granted upfront | Broad by default (send, readonly, compose, email) |
| Silent failure on first attempt? | Yes - scenario reported “completed,” email never sent | No | No |
| How the failure surfaced | Only by right-clicking the module and running it in isolation - the main log never showed it | N/A | N/A |
| Default trigger interval | Every 15 minutes | Every 1 minute | Not disclosed in our test (see note below) |
| Free-tier cap | 1,000 operations/month | 1,000 executions/month | 100 credits/day |
| Free tier exhausted by default polling alone | ~10 days | ~17 hours | N/A - hand-building in the editor consumed 0 credits (Sep 7); one AI Chat run used up the free credits (Sep 26) |
| Did the test email actually arrive? | Yes, once OAuth was fixed | Yes, confirmed real message ID | Yes, confirmed real message ID |
| Advertised AI builder actually present? | N/A (no AI builder claimed) | Yes, worked as advertised | Not on Sep 7, yes on Sep 26 - built a two-step flow from one sentence, but our free credits hit zero in that run, its first test email had Title and Link swapped, and we never saw the fix confirmed |
Note on Activepieces’ polling interval: unlike Make and n8n, Activepieces’ free tier charges credits only for published, live flow runs - not for anything done while building or testing in the editor. That meant the same “default interval silently burns your free tier” risk that mattered for Make and n8n wasn’t the pressing question for Activepieces, so we didn’t isolate a specific default-interval number in that test. Worth checking directly if polling frequency matters for your use case.
What this actually tells you
The failure that costs you real time isn’t “does it work,” it’s “does it tell you when it doesn’t.”
Make’s scenario reported success while a Gmail step silently failed - the only way to find the real
[403] insufficient authentication scopes error was to right-click the module and run it in isolation,
bypassing the trigger entirely. Neither n8n nor Activepieces reproduced this failure mode in our test,
though n8n’s assistant went a step further and proactively disclosed when a test run was simulated
rather than real, before we even asked.
“Free forever” doesn’t mean the same thing across these three tools. Make’s and n8n’s free tiers are consumed by the trigger’s own polling, whether or not anything new is found - meaning a workflow you publish and forget about can silently exhaust its allowance in hours (n8n’s default) to days (Make’s default). Activepieces’ credit model only charges for published runs, and building/testing costs nothing - a meaningfully different risk shape for anyone experimenting before committing to a live workflow.
A feature listed on a pricing page isn’t the same as a feature you can use today, and shipping it isn’t the same as it working. n8n’s AI assistant built our entire workflow from a single sentence, and it worked exactly as advertised. Activepieces listed the same category of feature (“Chat to automation,” labeled Beta) on its pricing page and documentation. On Sep 7 we searched the live product (sidebar, trigger search, global in-app search) and found nothing. We re-checked on Sep 26, after a promotional email said it had launched, and it was there. It built a two-step flow from one sentence, but in our one run the free credits hit zero, the first test email had Title and Link swapped, and the fix it announced was never confirmed. That is one run, not a verdict; see the Activepieces review for the screenshots.
Which should you use?
- Pick n8n if you want an automation built for you from a plain-English description, and you’re disciplined about checking the trigger interval immediately after - its default is the most aggressive of the three.
- Pick Make if you’d rather see and control every module yourself, and you’re comfortable digging into isolated module tests when something silently doesn’t run.
- Pick Activepieces if you want to build by hand and iterate without worrying about burning a free-tier allowance while you’re still building. Treat its new AI Chat as untested at scale: it exists now, but our one run used up the free credits.
Tested: Sep 2026 (Make Aug 11, n8n Aug 13, Activepieces Sep 7; Activepieces AI Chat re-tested Sep 26) · Test task: RSS feed → Gmail notification on a new item · Accounts: fresh, free-tier accounts for each tool · Evidence: screenshots and execution logs in each tool’s individual review, linked above.