StackReceipt

Automation guides / n8n

n8n buyer guide

n8n: count the starts, then name the owner.

A document request, a completion callback and a reminder can be separate executions. Work through the count, then choose Cloud or self-hosting based on who will operate it.

By Penny · StackReceipt · Documentation analysis, not a hands-on review

The short version

How n8n works

A useful candidate has a clear starting event, predictable inputs, and an outcome someone can check. “Keep two systems in sync” is too broad until you define which records, which direction, and what happens when they disagree.

Move data between systems.

Example to evaluate: an incoming order event starts a request to another system. Check that the destination API supports the exact operation and that your account can perform it.

Prepare information before a handoff.

Example to evaluate: take fields from an incoming record, standardize them, and send the result for review. Decide which missing or ambiguous values should stop the process.

Run recurring checks.

Example to evaluate: check a source on a schedule and route exceptions to an owner. Count scheduled starts as well as useful changes. These are possible jobs to investigate, not tested templates.

Fit & tradeoffs

n8n strengths and limitations

Our editorial judgment: consider n8n when someone can explain the data, inspect a failed request, and own the workflow after launch. A technical owner matters more than company size.

Reasons to consider it
  • An unusual API operation. The HTTP Request node allows a configurable REST request, rather than limiting evaluation to a named app action. That still requires endpoint and authentication knowledge. Source.
  • Deployment choice. Cloud and self-hosting let you decide who runs the infrastructure. Source.
  • A long workflow. A whole-run billing unit is worth comparing when many steps happen inside each run. It does not prove the lowest total cost. Source.
Reasons to pause
  • No one can maintain credentials, understand errors, or change the process when an upstream app changes.
  • You need a particular app action, but have only confirmed that the app has an integration logo.
  • Your decision assumes a free installation includes every collaboration or governance feature.
  • Your budget covers software but leaves no time for testing, recovery, or ownership.

These are evaluation criteria, not a measured ease-of-use ranking.

What will I need to learn?

Start with fields and records, the difference between an incoming event and an outgoing request, and how credentials grant access. For API work, practice reading an endpoint’s required inputs and error responses. Cloud removes server installation; it does not define your business rules for you.

Use one narrow workflow first. Explain what each step receives and returns before adding branches or more applications. We have not measured setup time or learning speed.

Worked example · hypothetical

One document process. Three kinds of workflow start.

Suppose a team sends supplier documents for signature. This design uses three independently triggered production workflows, not one workflow waiting for every stage. It assumes the signing service exposes the required API and completion webhook; no particular signing-service integration has been tested.

  1. Request receivedWebhook → validate supplier ID → create signature request
  2. Signing completedA separate webhook → match request ID → update the record
  3. Daily reminder checkSchedule Trigger → find unsigned requests → send due reminders
New requests
300 distinct inbound events = 300 top-level executions.
Completed signatures
240 distinct completion callbacks = 240 separate executions.
Daily reminders
One scheduled start per day for 30 days = 30 executions, not one execution per reminder message sent inside those runs.

300 + 240 + 30 = 570 executions

Assume all these runs complete normally, no callbacks repeat and no other workflows run. More nodes inside the same run do not multiply this count. Signature-service subscriptions, API fees, hosting and maintenance are separate; 570 is not a cost quote.

Execution accounting, Webhook and HTTP Request, checked . This is a modeled design, not an executed workflow.

Changing the design changes the count.

A callback that resumes a Wait node is not the separate Webhook-triggered workflow drawn above: Wait continues the paused execution. Likewise, n8n documents that called sub-workflows do not add to the parent’s execution quota. Do not count every box as a new run—or assume every stage of a business process is one run. Both rules checked .

The failure case: a completion arrives twice.

If ten valid completion callbacks are delivered again and start the webhook workflow, the model rises to 570 + 10 = 580 executions. The business record should still show only 240 distinct completed signatures. Store and check a stable request or event ID before repeated writes or notifications; that protection must be designed and tested.

Failed-run and retry billing are not established by this example. Before buying capacity for a failure-heavy workload, reconcile your account usage with the applicable plan terms.

Hosting & cost

Cloud or self-hosted?

The reviewed deployment guide separates managed Cloud from an installation on your infrastructure. Cloud has Starter, Pro, and Enterprise options; self-hosted options include Community, Business, and Enterprise. Source.

Cloud or self-hosted?

Cloud puts infrastructure management with n8n. Self-hosting puts configuration and maintenance with your team. Choose based on required control and operating capacity, then compare features. Business is a self-hosted option, so it is not simply the next hosted capacity tier.

Our checklist for self-hosting: name the owner for upgrades, credentials, backups, monitoring, and restores. Include infrastructure and staff time in the decision.

Is Community Edition the same as a paid plan?

No. The edition documentation excludes capabilities such as SSO, projects, workflow/credential sharing, and Git version control from Community Edition. Registration unlocks selected additional features; it does not turn Community into a paid plan. Check the exact edition against how your team will collaborate. Source.

How do executions affect the bill?

The execution documentation distinguishes trigger types. A Schedule Trigger firing every five minutes produces 30 × 24 × 12 = 8,640 starts in a 30-day month. A native polling trigger is different: empty polls do not count. Use the actual trigger type, not a generic “polling” label.

Manual, called sub-workflow and designated error-workflow executions are excluded from the quota by the documentation. Do not generalize that to every retry or failure. A monthly allowance also does not promise response time. Execution rules, checked .

What about the license?

The published Sustainable Use License limits use to internal business purposes or non-commercial/personal use and imposes distribution restrictions. Separately licensed enterprise files are excluded from that license. Public source code does not mean unrestricted resale or embedding rights.

If your business would offer n8n access or automation hosting to customers, get the intended arrangement confirmed against the applicable license and agreement before building around it. This guide does not determine your legal eligibility. License source.

Reliability & security

Handling failed runs

These are StackReceipt evaluation questions. They are not claims that every edition supplies a particular control.

How will we detect and recover a failed job?

Agree who is notified, what context they receive, and how they decide whether a retry is safe. Test an expired credential, a rejected input, a slow API, and a duplicate event. Compare the destination’s state with the execution record rather than assuming a green status means the business outcome is correct.

Will the debugging history still exist?

The pricing FAQ says saved execution history is constrained by record count, storage, and retention. Reaching a history limit removes older records without stopping workflows. Ask whether the available evidence will last through your incident-review cycle. Source.

What should security and enterprise buyers ask?
  • Who can view or change workflows, credentials, and execution data?
  • Which sign-in, provisioning, audit, and retention controls are included in this exact edition?
  • Where do payloads, logs, backups, and connected AI/API services send or retain data?
  • Who owns upgrades and incident response, and what does the support contract promise?
  • Can the team demonstrate access removal and restoration from backup?

A hosting location alone does not establish every downstream service’s data handling. Request evidence for your actual design and contract.

Beyond the monthly price

What happens when usage grows, a run fails, or another person needs access? These are documented limits; the checks below are our analysis, not results from product testing.

n8n

Many requests arrive at once

n8n Cloud limits simultaneous production executions separately from the monthly execution allowance. In regular mode, excess production executions wait in a first-in, first-out queue.

What to check

Estimate your busiest minute as well as your monthly total. A plan can cover the monthly count while making a burst of webhook requests wait. Test whether that delay is acceptable.

Read the vendor documentation · Checked

n8n

Someone else needs to manage the workflow

Self-hosted Community Edition excludes Projects, workflow and credential sharing, and SSO. Free registration unlocks some additional features, but does not remove those collaboration restrictions.

What to check

For a shared production workflow, compare the relevant paid edition as well as server costs. Free self-hosting is not a substitute for every paid collaboration feature.

Read the vendor documentation · Checked

Before subscribing or self-hosting

Prove the boundary and the handover.

Use synthetic supplier IDs and a sandbox destination. The checks below are a proposed acceptance test, not results we observed.

  1. Draw every start. Mark independent webhooks, schedules, native polling triggers, called sub-workflows and Wait resumptions. Reconcile those boundaries with the execution list.
  2. Repeat a completion event. The same request ID must not create a second completion notice or destination record. Inspect the destination, not just the workflow status.
  3. Fail after the external write. Simulate a timeout after a request is accepted. The operator must discover whether it succeeded before sending it again.
  4. Test the busy minute. Send the expected burst, then compare queued requests and completion time with the business deadline. A monthly quota is not a concurrency test.
  5. Hand it to the backup owner. Have another authorized person find a failed execution and identify required credentials. For self-hosting, also demonstrate a restore and document the upgrade owner.

Stop the self-hosted rollout if no one owns recovery, or if the chosen edition cannot provide the sharing and access controls the handover needs. Compare Cloud or an appropriately licensed edition before treating free software as the complete solution.

Common questions

Questions before buying

Can I use it without writing code?

The example above uses documented configurable nodes. That does not mean every workflow avoids code or technical concepts. Pick your actual operation and see whether the available fields express it; the HTTP Request documentation explicitly assumes API understanding. Source.

Does self-hosting remove all costs?

No. Treat the free Community edition separately from infrastructure, maintenance, third-party API charges, and any paid features you need. Use the edition comparison to find gaps before budgeting.

Can it work with an app that has no dedicated node?

A documented REST API may provide a route through HTTP Request. It is not a compatibility guarantee: confirm the operation, authentication, permissions, payload, and service limits. An inaccessible or unsupported API operation remains a blocker. Source.

When should I compare Zapier, Make, or a direct script?

Compare the same job in Zapier and Make if their documented app actions meet your requirements. If you already maintain a small script for the job, include its ownership and debugging effort too. Judge the completed task, recovery work, and total cost; this guide does not establish a universal winner.

What has StackReceipt actually verified?

We inspected the linked documentation and separated its rules from our hypothetical design. We did not connect accounts, execute workflows, measure reliability, or verify your entitlement. Retry and failed-run billing still need account-specific verification; source dates below distinguish new checks from earlier research.

Put your own volume into the comparison.
The calculator is a simplified estimate; check its assumptions.
Compare your workload →

Sources and limits

The 570-execution example is arithmetic for the stated three-workflow design, not measured usage or a working template. No signing-service endpoint, account permissions or current subscription prices were verified. Dates below belong to individual source checks; the buying notes retain their own earlier observations.

Fit, architecture and trial recommendations are StackReceipt editorial judgment, not a reliability benchmark or legal determination. Research method · Guide index and billing terms.