n8n vs Make

Quick answer

Updated October 27, 2026. n8n and Make.com are both visual workflow builders, but n8n can run on your own servers, bills its cloud plans per workflow execution with unlimited steps, and ships native Evaluations for AI output.[27, 13] Make.com is a hosted service billed in credits that usually follow module operations, and it advertises 3,500+ apps against n8n's 2,000+ integrations.[31, 45, 75] For long processes, note that Make.com's Sleep module pauses at most 300 seconds, while n8n's Wait node saves the run and resumes it later.[43, 12]

Key takeaway

Choose n8n when you need to self-host, when runs must pause for hours or days, or when you want to score AI output against a test dataset.

Choose Make.com when you want a hosted builder with the larger advertised app catalog and a free plan, and your scenarios finish within 40 minutes.

  • n8n wins on self-hosting: a self-hosted Community Edition, a self-hosted Business plan, and hosted or self-hosted Enterprise.
  • Make.com wins on advertised app catalog: 3,500+ standard apps on every plan, against n8n's 2,000+ integrations.
  • n8n wins on long waits inside a run: the Wait node stores the run in the database and resumes it by time, webhook or form.
  • n8n wins on testing AI output: native Evaluations run datasets through a workflow and score the results.

By , CTO, Trampoline AI. Updated · Facts checked · Published by the Avalanche team. Avalanche competes with both.

01 / The core differencen8n vs Make.com

Where it runs and how long a run may last

Both products let you draw a workflow, connect apps and add AI agents. They differ in where the platform runs, how long one run may last, how a failed run recovers, and what you pay for. Those four points decide most n8n versus Make.com evaluations, so this page covers each with the vendors' own documentation.

n8n is a platform you can operate yourself, with a durable Wait node and a configurable workflow timeout when self-hosted.[12, 35] Make.com is a hosted service where one scenario run stops at 5 minutes on Free and 40 minutes on paid plans, and longer waits are split into separate scenarios joined by a data store.[31, 43] n8n Cloud also caps a run, at 5 minutes on Starter and 40 on Pro and Enterprise.[27]

02 / At a glanceChecked 2026-10-27

n8n and Make.com, side by side.

n8n and Make.com compared, checked 2026-10-27
Criterionn8nMake.comEdge
Self-hostingCommunity Edition, Business (self-hosted) and EnterpriseHosted only; Enterprise on-prem agent connects local apps[27, 29]n8n edge
Advertised app catalog2,000+ integrations3,500+ standard apps[75, 31]Make.com edge
Billing unitWorkflow executions, unlimited steps eachCredits, usually one per module operation[27, 45]Depends
Long waits inside a runWait node saves the run and resumes by time, webhook or formSleep pauses 300 s at most; longer waits need a second scenario[12, 43]n8n edge
Retry after a failureRetry keeps completed nodes and resumes the failed oneStored incomplete executions resume at the failed module[11, 22]Similar
Testing AI outputNative Evaluations with datasets and metricsManual: Run once, agent chat and replay[13, 14, 9, 21]n8n edge
Version history1 day Starter, 5 days Pro, 365+ days EnterpriseEach saved version kept 60 days[27, 6]Depends

Catalog sizes are each vendor's own count on October 27, 2026; an n8n integration and a Make.com app are not the same unit. Make.com does not document a scored evaluation feature in the pages reviewed, which is not proof it lacks one.

03 / Where each one winsOne claim per card

Where each one wins

Where n8n wins

  • Runs on your servers

    The Community Edition and the Business plan are self-hosted, and Enterprise can be hosted or self-hosted.[27]

  • Waits that survive

    The Wait node saves the execution to the database and resumes it after an interval, at a date, or on a webhook or form submission.[12]

  • Scored AI tests

    Evaluations run a dataset through a workflow, compute deterministic, AI or custom metrics, and compare runs case by case.[13, 14]

  • Approval before a tool runs

    Human review pauses an AI agent before a selected tool call; Approve runs it and Deny cancels it.[16]

  • Git environments

    Business and Enterprise link instances to git branches, so a change can move from development to production through a pull request.[3]

Where Make.com wins

  • More ready-made apps

    Make.com advertises 3,500+ standard apps on every plan, including Free.[31]

  • A real free plan

    Free includes 1,000 credits a month, routers and filters, and two active scenarios, with a 15-minute minimum schedule interval.[31]

  • Build by chat

    Maia builds and edits scenarios and agents from a conversation; it is in public beta on paid plans, with a trial on Free.[70]

  • Per-module run detail

    Scenario history shows each module's bundles, output, status, duration and credits, and exports to CSV.[44]

  • Capacity in view

    The parallel capacity dashboard charts each scenario's use of the organization's processing units.[25]

04 / Detailed comparison4 buyer questions

What changes for your team.

01

How long a run can last

Buyer question

Can n8n or Make.com run a process that takes hours or waits for a reply?

Make.com's pricing lists a maximum scenario execution time of 5 minutes on Free and 40 minutes on Core, Pro, Teams and Enterprise.[31] The Sleep module delays a run for up to 300 seconds; for longer waits, Make.com's own guidance is two scenarios: one writes a timestamped record to a data store, a later scheduled scenario picks it up.[43] Make AI Agent (New) adds a per-step timeout that defaults to 300 seconds and caps at 600.[21]

n8n Cloud caps an execution at 5 minutes on Starter and 40 minutes on Pro and Enterprise; self-hosted instances set their own timeout.[27, 35] The Wait node persists the execution to the database and resumes it after an interval, at a date, or when a webhook or form arrives, although waits under 65 seconds stay in memory.[12] If your process waits on a person or another system for days, n8n models that in one workflow and Make.com models it as two.

Sources [12, 21, 27, 31, 35, 43]

02

When a run fails, and when you need to roll back

Buyer question

Do n8n and Make.com rerun the whole workflow after an error?

Neither has to. With execution progress saved, an n8n retry keeps the results of completed nodes and resumes at the failed node, using either the original or the current workflow; the failed node reprocesses all of its items.[11] With Store incomplete executions turned on, a Make.com retry starts at the failed module with the configuration it had at failure.[22] Make.com's separate Replay feature is different: it runs the current scenario from old trigger data through every module, so it can repeat side effects.[9]

For versions, n8n can restore a past version to the draft or publish any version from history; history lasts 1 day on Starter, 5 on Pro and 365+ on Enterprise.[1, 27] Publishing is not all-or-nothing: n8n documents a "Published, partial" state where some triggers activate and others fail.[1] Make.com keeps each saved version for 60 days; you select it, choose Restore version, then Save, because restoring alone does not persist.[6]

Sources [1, 6, 9, 11, 22, 27]

03

Executions versus credits

Buyer question

Is n8n cheaper than Make.com?

As of October 27, 2026, n8n Starter is €20 a month billed annually for 2,500 workflow executions, and Pro is €50 a month billed annually for 10,000, each with unlimited steps.[27] Make.com shows Core at $9, Pro at $16 and Teams at $29 a month at the 10,000-credit tier, each marked "Save 15%", and Free at 1,000 credits a month.[31]

The units do not convert. One n8n execution covers every node in the run. In Make.com most non-AI module operations use one credit, a module runs once per bundle it processes, AI features can bill by tokens, and the Make Code app uses 2 credits per second of runtime.[45, 31] A ten-module scenario that processes 50 rows can use hundreds of credits in a single run, so price the same workflow on both before deciding.

Sources [27, 31, 45]

04

AI agents, testing and review

Buyer question

Which is better for AI agents, n8n or Make.com?

n8n's AI Agent node runs a tool-calling agent built on LangChain JavaScript, with model, tool and output-parser sub-nodes.[66, 67] Human review can stop the agent before a chosen tool runs, and Evaluations score agent output against a dataset, including cases taken from production failures.[16, 13] n8n's newer standalone Agents are in Preview and do not support queue mode yet.[15]

Make AI Agent (New) runs inside a scenario with mapped inputs, module, scenario and MCP tools, knowledge files and structured output; it is in open beta.[21] Make.com suggests adding an output-review tool and a Slack confirmation for human sign-off, which is guidance rather than an engine-enforced gate.[24] Teams that need to test agents systematically will find more built in on the n8n side today.

Sources [13, 15, 16, 21, 24, 66, 67]

05 / FAQDirect answers

Common questions.

Can Make.com be self-hosted?
No self-hosted edition is documented. Make.com runs as a hosted service; its Enterprise on-premise agent lets scenarios reach apps on a local network, but the platform itself stays hosted.[29, 31] n8n offers a self-hosted Community Edition and self-hosted paid plans.[27]
What is the maximum run time on Make.com?
One scenario execution can run 5 minutes on Free and 40 minutes on paid plans, and the Sleep module pauses at most 300 seconds.[31, 43] n8n Cloud has the same 40-minute cap on Pro and Enterprise, but its Wait node can park a run and resume it later.[27, 12]
Is n8n open source?
n8n is source-available under the Sustainable Use License, which restricts some commercial uses; n8n says it is not open source by the OSI definition.[38]
Does Make.com still charge operations?
Make.com bills credits. An operation is one module run, and most non-AI operations cost one credit; AI, file and code features can cost more or bill by usage.[45]
Does n8n or Make.com have human approval for AI agents?
n8n documents human review that pauses before selected AI tool calls, with approve or deny through chat, Slack, Telegram and other channels.[16] Make.com documents a review pattern built from an output-review tool and a Slack message.[24]
06 / Where Avalanche fitsNeither of these

Where Avalanche fits

Avalanche is a third option for long, document-heavy agentic workflows that outgrow a 40-minute run: you write a few coarse steps in Python and agents work inside each one. On Avalanche Cloud, every step is a checkpoint, so a failed step retries without redoing finished steps, and each run's typed step inputs and outputs stay in one history for your organization, so you can compare a step across many runs and versions. Schedules and webhooks are declared in the workflow code and follow rollbacks. The runtime is MIT-licensed.

See how Avalanche works
07 / SourcesChecked 2026-10-27

Public primary sources.

Every source was fetched on the date shown. Prices and plans change; check the vendor before you buy.

  1. [1]Save and publish workflowsn8n · Checked https://docs.n8n.io/build/understand-workflows/save-and-publish-workflows
  2. [3]Source control and environmentsn8n · Checked https://docs.n8n.io/administer/use-source-control-and-environments
  3. [6]Scenario version controlMake · Checked https://help.make.com/scenario-version-control
  4. [9]Scenario run replayMake · Checked https://help.make.com/scenario-run-replay
  5. [11]View all executionsn8n · Checked https://docs.n8n.io/build/understand-workflows/understand-executions/view-all-executions
  6. [12]Wait noden8n · Checked https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.wait
  7. [13]Why test AI workflows (Evaluations)n8n · Checked https://docs.n8n.io/build/integrate-ai/test-and-improve-ai-workflows/understand-why-to-test
  8. [14]Use metrics to measure qualityn8n · Checked https://docs.n8n.io/build/integrate-ai/test-and-improve-ai-workflows/use-metrics-to-measure-quality
  9. [15]Build and manage agentsn8n · Checked https://docs.n8n.io/build/build-and-manage-agents
  10. [16]Human-in-the-loop for toolsn8n · Checked https://docs.n8n.io/build/integrate-ai/ai-examples/human-in-the-loop-for-tools
  11. [21]Make AI Agent (New) appMake · Checked https://help.make.com/make-ai-agent-new-app.md
  12. [22]Manage incomplete executionsMake · Checked https://help.make.com/manage-incomplete-executions.md
  13. [24]Make AI Agent (New) best practicesMake · Checked https://help.make.com/make-ai-agent-new-best-practices
  14. [25]Parallel capacityMake · Checked https://help.make.com/parallel-capacity
  15. [27]n8n Plans and Pricingn8n · Checked https://n8n.io/pricing
  16. [29]On-premise agentMake · Checked https://help.make.com/on-premise-agent
  17. [31]Make Pricing & Subscription PackagesMake · Checked https://www.make.com/en/pricing
  18. [35]Configure workflow timeoutsn8n · Checked https://docs.n8n.io/deploy/host-n8n/configure-n8n/basic-configuration/configuration-examples/configure-workflow-timeouts
  19. [38]n8n Community licensen8n · Checked https://docs.n8n.io/n8n-community-license
  20. [43]Tools (Sleep module)Make · Checked https://help.make.com/util
  21. [44]Scenario historyMake · Checked https://help.make.com/scenario-history
  22. [45]CreditsMake · Checked https://help.make.com/credits
  23. [66]Tools Agentn8n · Checked https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/tools-agent
  24. [67]LangChain in n8nn8n · Checked https://docs.n8n.io/build/integrate-ai/langchain-in-n8n
  25. [70]Introduction to Maia by MakeMake · Checked https://help.make.com/introduction-to-maia-by-make
  26. [75]n8n integrations directoryn8n · Checked https://n8n.io/integrations
Related comparisonsView all comparisons
See the runtime for yourselfOpen source · MIT

Our code speaks for itself.

The Avalanche runtime is open source under the MIT License. Read the code, run the starter project, and see how it fits the work your team repeats.