Make vs n8n: Credits per Step or One Execution per Run
Both draw automations as boxes on a canvas. The difference is what they count and where they run. Make bills credits for the modules a run touches and runs only in its cloud. n8n bills one execution per whole run and can also run on your own server. This page counts the same workloads in both units.
By Elias Brandt · Updated 8 Oct 2026 · About 5 min read · Plan terms taken from each tool profile
The short answer
Choose Make for a hosted visual workflow and a lasting Free plan when its credit allowance covers your actual traffic. Choose n8n when self-hosting or its execution-based meter fits your operations. Both support JavaScript and Python. Compare monthly runs, billable processing steps and hosting work before choosing; a four-module workflow is not a price break-even point.
Make or n8n: which one, when?
Pick by how long a run is and where it must live
In shortShort runs that start from a webhook favour Make; long runs, heavy volume or in-house hosting favour n8n.
Make for short, event-driven scenarios
When most runs touch a handful of modules and start from a webhook rather than a schedule, credits stretch furthest, and a light process can stay on the Free plan indefinitely.
n8n for long runs at volume
When one run chains ten or more steps or calls several AI models, per-run billing keeps the count flat: Pro's 10,000 executions allow about one run every 4 minutes all month (43,200 minutes / 10,000).
Self-hosted n8n when data must stay in-house
Community costs no licence fee but needs someone who can update and restore the instance; Business adds SSO, SAML and LDAP at EUR 667 a month billed annually.
Six everyday jobs, side by side
In shortRead across a row: the job, then what each tool does and what it counts.
| If the job is… | Make | n8n |
|---|---|---|
| Poll an app every 15 minutes | About 2,880 checks a month, each using a credit even when nothing arrives: nearly three times Free's allowance | About 2,880 executions a month, 380 more than Starter includes and under a third of Pro |
| Start on an incoming request | Custom or app-specific webhook with its own request queue, so no empty checks are billed | Trigger node starts the workflow; the whole run counts once |
| Route records by a condition | Routers and filters with a fallback route for unmatched data; each module on the path uses credits | Node-based branches inside one workflow; still one execution per run |
| Add custom logic | JavaScript or Python through paid Make Code; 2 credits per second, with custom imports on Enterprise | Code node in JavaScript or Python; library and execution restrictions depend on hosting |
| Recover from a failed step | Skip, Retry, Resume, Commit or Rollback handler on the failing module | Error workflow starting with an Error Trigger node, plus retry from the execution list |
| Keep run logs | 7 days on Free; paid search and retention depend on the selected package | 7 days on Starter, 30 days on Pro |
The basics
In shortRead live from each profile.
| What do you need? | Make | n8n |
|---|---|---|
| In short | Connect app modules in visual automation scenarios. | Build workflows in managed Cloud or on your own infrastructure. |
| Free access | Free access available | Free access available |
| Plans |
|
|
| Price check | Pricing review due · Sep 25, 2026 | Pricing review due · Sep 24, 2026 |
| Full profile | Make profile | n8n profile |
Values come from each tool’s profile and update with it, in the currency each vendor shows.
Visit Make (opens in a new tab) Visit n8n (opens in a new tab)
How is each one built, and where does it run?
In shortMake is a hosted service only. n8n is the same product whether n8n hosts it or you do.
Six differences that decide it
- Billing unit
- Make charges credits for module runs, normally one per ordinary module and again for every bundle a module processes; n8n counts one execution per complete workflow run, however many steps it has. Source: help.make.com
- Worked example
- Assuming ordinary modules that each handle one bundle, a 10-step workflow run 100 times a day uses about 30,000 Make credits a month (10 x 100 x 30), three times an illustrative 10,000-credit budget, but only about 3,000 n8n executions, which fits within Pro's 10,000. Source: n8n.io
- Free access
- Make Free never expires: 1,000 credits a month, 2 active scenarios and a 15-minute minimum interval. n8n Cloud has only a 14-day, 1,000-execution trial; its lasting free route is the self-hosted Community edition. Source: make.com
- Hosting
- Make runs only as a hosted browser service. n8n runs as managed Cloud or on servers you operate, where upkeep and restores are your job. Source: docs.n8n.io
- Entry quote
- Make public pricing views show different plan names and credit packages; confirm a current quote. n8n Starter lists EUR 20 per month on annual billing for 2,500 executions. Match volume, currency and commitment before comparing. Source: make.com
- Developer depth
- Both tools execute JavaScript and Python. Make Code requires a paid plan and meters runtime; n8n library access depends on Cloud versus self-hosted deployment. Source: help.make.com
Feature matrix
In shortFourteen capabilities marked from the two profiles. A dash means the profile does not cover it, not that the tool lacks it.
| Capability |  Make |  n8n |
|---|---|---|
| Building a workflow | ||
| Visual canvas | Offeredscenario builder | Offerednode-based canvas |
| Start from a template | Offeredpublic library, 7,500+ use cases | Offeredtemplate library; self-hosted can serve its own |
| Write code in a step | Partlypaid Make Code, 2 credits a second | OfferedCode node; limits depend on hosting |
| Call any REST API | Not listednot covered in the sources checked | OfferedHTTP Request node, cURL import |
| Build a workflow from a chat prompt | PartlyMaia, beta, paid plans | Not listednot covered in the sources checked |
| AI | ||
| AI agents | PartlyAI Agent app, open beta | Offeredseveral model providers, tools and memory |
| MCP | OfferedMCP server: Claude or ChatGPT run your scenarios | OfferedMCP servers as tools for an agent |
| Running it | ||
| Start instantly from a webhook | Offeredown request queue per webhook | Offeredtrigger node |
| Handle a failed step | OfferedSkip, Retry, Resume, Commit, Rollback | Offerederror workflow, retry from the list |
| Runs in parallel | Not listednot covered in the sources checked | Offered5 on Starter, 20 on Pro |
| Hosting and control | ||
| Run on your own server | Not listedcloud only | OfferedCommunity edition, no licence fee |
| Single sign-on | Not listednot covered in the sources checked | PartlyBusiness: SSO, SAML, LDAP |
| Git-backed environments | Not listednot covered in the sources checked | PartlyBusiness and Enterprise |
| Price per seat | Not listednot covered in the sources checked | Offeredno seat pricing; unlimited users |
offered partly or with limits not listed on the profile.
What does a month cost, counted in each tool’s unit?
In shortMake counts every module and every bundle a module handles; n8n counts a whole run once. Count your load first, then ask for prices.
The units, row by row
| Work | Make | n8n |
|---|---|---|
| One ordinary step handling one record | 1 credit | Included in the run’s single execution |
| The same step handling 50 records in one run | About 50 credits: one per bundle | Still one execution |
| A scheduled check that finds nothing | 1 credit | 1 execution |
| Code inside a step | 2 credits per second of runtime, paid plans | Included in the run; limits depend on hosting |
| AI through your own provider account | Make credits plus the provider’s token bill | Still one execution per run; provider billing is not covered in the sources checked |
Units as each profile states them. Bundle counting assumes ordinary, non-AI modules.
Sources: Make credits, Make operations and bundles, n8n pricing.
Three monthly loads, counted
In shortSame work, two meters. A 30-day month.
| Load | Make credits | n8n executions |
|---|---|---|
| One check an hour, single step | 720 (24 × 30): leaves about 280 of Free’s 1,000 | 720: inside Starter’s 2,500 |
| Webhook brings 300 records a day; four steps each | About 36,000 (4 × 300 × 30) | 9,000 (300 × 30): inside Pro’s 10,000 |
| Same 300 records, batched into one nightly run | About 36,000: every record still counts per step | 30: one run a night |
Arithmetic on the per-module and per-run rates above, assuming one record per bundle and ordinary modules. Batching changes the n8n count, not the Make count.
What each plan includes
| Make plan | What it includes |
|---|---|
| Free · USD 0 | 1,000 credits a month, 2 active scenarios, checks every 15 minutes at most |
| Paid automation | Confirm a current quote: the public pricing pages show different plan names and credit packages |
| Enterprise | Custom pricing; advanced security, custom libraries and overage protection |
Price per month on yearly billing, from the n8n profile. Bars compare plans of this tool only.
n8n Cloud prices are monthly equivalents on annual billing, in euros; Business, self-hosted, is in the table further down. Make’s paid price is not compared here because its public pages disagree; count your credits, then ask Make for a quote in the same billing period.
Self-hosting n8n: what do you take on?
In shortThe licence can be free; the server, the updates and the restores are yours.
Who runs what
| Make (cloud) | n8n Cloud | n8n self-hosted | |
|---|---|---|---|
| Servers and updates | Make | n8n | You |
| Restore after a failure | Make | n8n | You |
| Licence | Credit plan or quote | EUR 20 or 50 a month, annual | Community: none · Business: EUR 667 a month, annual |
| Single sign-on | Not covered in the sources checked | Not listed for Starter or Pro | Business: SSO, SAML, LDAP |
| Moving work between environments | Not covered in the sources checked | Not listed for Starter or Pro | Business: Git-backed environments |
| If nobody on the team can run a server | Fits | Fits | Start on Cloud instead |
From the n8n deployment and pricing pages and the Make plan list on each profile. Make’s profile says to ask for its deployment and private-network options under contract before choosing it for infrastructure you must control.
Sources: Choose how to use n8n, n8n source control and environments, Make pricing.
Videos, questions and what this page does not judge
Official videos
In shortFrom each vendor’s YouTube channel. They load only when you press play.
Common questions
Which is cheaper, Make or n8n?
Build a monthly bill from your real run volume before choosing. Include bundles, code runtime, AI-provider charges and any higher credit package on Make; include the required execution tier or server maintenance on n8n. Convert currencies on a stated date and match billing periods. Four modules is a quota calculation, not a universal price threshold.
Does one Make credit equal one n8n execution?
Use a billable-work count rather than the number of boxes drawn in the editor. Routers and error handlers can be credit exceptions on Make, while code and AI operations can use a different rate. Check run history with your actual record batches; an estimate based on one record will undercount a fan-out over many records.
Do empty scheduled checks cost money on both?
Yes. A Make scheduled trigger uses a credit every time it looks for data, even when nothing is waiting, and an n8n scheduled workflow uses one execution per run. An hourly check costs about 720 of either unit a month; switching Make to a webhook removes those empty checks.
What happens if I go over the limit?
n8n keeps workflows running when a quota is exceeded and invoices unresolved overages automatically 45 days later; on Business, extra executions come in buckets of 300,000 for EUR 4,000. Make lists overage protection as an Enterprise feature, so on lower tiers plan headroom or a higher credit tier.
Can I start on the n8n Cloud trial and move to self-hosting later?
Yes, but export first. The trial includes Pro features, 1,000 executions and no card requirement; if it is not converted, the workspace closes and its workflows stay downloadable for 90 days. Self-hosted Community then runs them without a subscription, on a server you maintain.
What this page does not judge
Which editor is easier to learn, how fast each runs, and how many app connectors each offers are the topics most comparisons rank on, but neither profile documents them in a way we can check. We have not built the same workflow in both, so this page compares what each one counts and where it runs. The test below measures it on your own workflow.
Take one workflow you already run by hand, with a trigger, three steps and about 50 records. Build it on Make Free and on the n8n Cloud trial. Run it once with real data, then read the credits used in Make’s execution history and the executions used in n8n’s list, and note any step you could not build in either.
How this comparison was made
Plans, units and limits come from the Make and n8n profiles on this site, which cite the vendors’ pricing, help and documentation pages and carry their own check dates; tables marked as live are read from those profiles each time the page loads. Sections follow the topics the top-ranking pages cover, limited to what the vendors document. No rating is given. See our editorial policy.
