What does n8n Starter include in October 2026?

The cited n8n plan lists 2.5K workflow executions, unlimited steps and five concurrent executions [1]. Those three limits answer different questions: how many runs are included, whether a run may have multiple steps, and how many can operate simultaneously. Put the firm's actual run pattern beside all three before comparing a plan with a Zapier task allowance.

What does Zapier Professional add for multi-step work?

Zapier Professional lists multi-step Zaps, unlimited Premium apps and webhooks [2]. That makes it a relevant plan for a workflow that needs those features, but access to them is not the same as an estimate of monthly task use. The trial should count the actions produced by a typical case and test a failed case before the business commits.

Are an n8n execution and a Zapier task the same unit?

No. The comparison starts with an n8n workflow execution and a Zapier task as differently defined usage units. A multi-step run may generate a different number of billable events across platforms, so the business should trace one real workflow through each plan's published usage definition [1][2]. Do not compare the printed allowance numbers without translating the actual case into each unit.

Can n8n be self-hosted without a licence key?

Yes. n8n's hosting documentation says the free Community edition runs without a licence key, while keys enable Business or Enterprise editions [3]. That describes edition access, not a zero-cost operating service. A company choosing self-hosting still needs to assign ownership for upgrades, credentials, monitoring, backups and failures before it treats the option as economical.

Does Zapier Professional remove every workflow limit?

The cited pricing page identifies multi-step Zaps, unlimited Premium apps and webhooks for Professional [2]. It does not follow that every possible workflow has unlimited monthly use or no other plan conditions. The team should inspect the current plan's task allowance and terms at purchase, then use its own test runs to estimate demand. A feature list alone cannot settle a subscription choice.

What should the trial show before choosing either tool?

It should show the number of real cases processed, the usage unit consumed by each case, elapsed time, failed-run behaviour and the person who restores service. A field-services team can test one job request from intake to dispatch confirmation. The same inputs and expected result should be used on both tools, with the plan limits documented beside the outcome [1][2][3].

n8n vs Zapier: Compare Workflow Limits and Hosting in 2026

Quick Answer: In October 2026, choose n8n vs Zapier: use self-hosted n8n Community when hosting control matters and your team can maintain it ³; use Zapier Professional when multi-step Zaps, Premium apps and webhooks fit ². n8n Starter includes 2.5K executions and unlimited steps ¹. Compare tasks with runs and test recovery before committing.

Summary in a mind map

n8n vs Zapier: Compare Workflow Limits and Hosting in 2026
├─ Select a case
│   ├─ One recurring field-service request
│   └─ Include duplicate and missing-data cases
├─ Count use
│   ├─ n8n counts workflow executions
│   └─ Zapier counts successful-action tasks
├─ Check access
│   ├─ n8n Starter: unlimited steps
│   └─ Zapier Pro: apps and webhooks
├─ Own operations
│   ├─ Community edition has no licence key
│   └─ Assign upgrades and recovery
└─ Decide
    ├─ Same inputs in both trials
    └─ Keep the offer and renewal terms

Choose one workflow to compare

Choose n8n if self-hosting and direct operational control are required and the team can maintain the service ³. Choose Zapier Professional if its multi-step, Premium app and webhook access meets the required route ². For either choice, measured usage and recovery effort must fit the current plan ¹^,2. The decision should begin with work the business already performs, not a catalogue of possible integrations. A Salt Lake City field-services company can choose one recurring case: a job request enters a form, a coordinator checks the service area, a dispatcher receives a task, and the customer gets a confirmation. Each stage needs a source system, expected output and owner. That makes the trial comparable across tools.

The company should prepare three test inputs: a normal request, a duplicate and a request missing a service address. The normal case tests the intended route. The duplicate checks whether a second run produces unwanted work. The missing address tests the failure path. These are proposed test cases for the firm, not claims that either plan supplies the business logic automatically.

A connection list follows from that workflow. The team should name the form, customer record, dispatch queue and message channel it actually uses. If a required system is absent from a plan's accessible connections, a lower usage allowance will not rescue the design. Conversely, a long integration list has little value if the four required connections work and the trial meets the service standard.

Compare executions with tasks

n8n Starter includes 2.5K workflow executions with unlimited steps and five concurrent executions ¹. Its published unit is a workflow run, while step count and concurrency are separate dimensions. Zapier presents task-based usage in its pricing framework ². A person comparing 2.5K executions with a task allowance needs a conversion based on the exact workflow, because the printed numbers do not share a unit.

Zapier's pricing FAQ counts successful app actions as tasks but excludes built-in Filter, Formatter and Path steps ². n8n says its pricing is based on monthly workflow executions regardless of complexity ¹. For a route with several successful actions, those definitions can produce different usage totals from the same customer request. The trial should record both the complete workflow run and the actions it completed, then apply each plan's current rules to those observations.

For the field-services case, trace one request from trigger to confirmation. Mark each stage that causes a platform-counted event under the current plan rules. Repeat for the duplicate and error cases. The resulting worksheet should contain measured use per case and the number of cases the business expects, clearly labelled as its own forecast. It should not call a test count a vendor guarantee.

Comparison	n8n Starter	Zapier Professional
Usage allowance	2.5K workflow executions ¹	Task allowance to check for the selected plan ²
Feature access	Unlimited steps ¹	Multi-step Zaps, Premium apps and webhooks ²
Other constraint	Five concurrent executions ¹	Count billable successful app actions, excluding built-in Filter, Formatter and Path steps ²
How to compare	Count complete workflow runs	Count billable tasks from the same case

For a hypothetical 100 requests, suppose each request produces one complete run and three billable successful app actions, with no retries. That is 100 n8n executions under its run definition ¹ and 300 Zapier tasks under its action definition ². Built-in Filter, Formatter and Path steps would add no Zapier task in this example ². The figures illustrate a conversion, not a quoted plan allowance.

A monthly forecast also needs peak demand. Five concurrent executions is relevant if many job requests arrive together, even where total monthly runs look modest. A slow downstream system can leave work in progress, so the team should observe peak behaviour in its trial. This is an operating test, not an assertion that the published monthly allowance predicts queue behaviour.

Check plan access before building

Zapier Professional includes multi-step Zaps, unlimited Premium apps and webhooks ². Those are feature access statements. The team should check whether its job-request route actually needs more than one step, a Premium connection or an incoming webhook. If none is needed, the presence of those features is not itself a reason to buy that plan. If one is essential, test it with the real workflow shape.

n8n Starter's unlimited steps remove a step count from that plan's listed limit, but do not remove the execution or concurrency limits ¹. A workflow with numerous steps may still be expensive to maintain if each condition is poorly documented. The trial should inspect the visible sequence, not merely ask whether the platform permits another node. The person who will repair it should be able to explain the route without its builder present.

Need in the trial	Evidence to capture
Multiple actions after one request	Whether the selected plan exposes the required steps ¹^,2
An external event starts the route	Whether the required webhook path works ²
Several requests arrive at once	Runs in progress against n8n's concurrency limit ¹
A required business app	Connection and permission test on the selected plan

The business should preserve screenshots or an export of the tested configuration, with sensitive credentials removed. That record helps distinguish a platform limit from a configuration error. It also gives the next maintainer a starting point after staff change.

Decide on n8n hosting

n8n documents a free Community edition that can run without a licence key. Its hosting page says licence keys enable Business or Enterprise editions ³. That distinction is about editions and licence access; it does not say that self-hosted operation is without labour, infrastructure or support cost. The company's purchase question should therefore include who will run the service.

If the IT desk hosts n8n, it owns the operational path: applying upgrades, protecting credentials, observing failed runs, maintaining backups and recovering a workflow after a broken connection. Those are work assignments the company must make. A hosted plan changes who operates the platform, but the company still owns the correctness of its business route and the data it sends through it.

The trial should include a maintenance day alongside a successful demonstration. Record where secrets are held, how a departed employee's access is removed, where failure notices go and who can pause a malfunctioning route. These questions decide whether self-hosting is a viable choice for this particular team. A licence key statement cannot answer them ³.

Assign failure and recovery owners

A workflow remains unfinished when only the normal case passes. The duplicate request might create two dispatch tasks; the missing address might leave a customer without confirmation. Assign a person to review each failure type and a route to correct it. The record should say whether a failed case can be retried safely or must be handled manually.

The team should decide where errors become visible. If an automated route fails silently, the apparent subscription cost hides manual investigation and missed service work. In the trial, intentionally break a non-production connection and note the time until a person notices and restores it. That observation is more useful than a generic claim that one product is reliable.

Credential ownership belongs in the same decision. A route connected through one employee's personal account can fail when that account changes. The business should identify the authorised account for each connection and the procedure for renewing it. This is a governance choice for the firm, not a feature promise from either pricing page.

Run the same trial on both

One week of testing can be structured around the same set of sample job requests, provided it represents the company's actual pattern. The team should record normal results, duplicates, failures, usage units and human time spent repairing cases. It should then calculate an operating-cost estimate using the current plan offer and its own expected volume. The cited pages identify plan features and n8n's editions; they do not supply the firm's demand ¹^,2^,3.

The result is a decision sheet, not a universal winner. If one tool handles the required systems with less recovery work, that may outweigh a difference in printed allowance. If self-hosting is essential, n8n's documented Community route is directly relevant ³. If a team values Professional's multi-step, Premium app and webhook access, it should test those exact needs on Zapier ².

What the comparison cannot settle

These sources do not fix the field-services firm's monthly task count, infrastructure cost or renewal terms. A public plan page can change, and a business's actual commitments may depend on the order it accepts. The team should retain the offer and terms shown at purchase. It should not infer an annual commitment or cancellation right from a feature bullet.

The surrounding decisions differ from choosing an automation platform. Read AI agents versus automation platforms for the system boundary, customer support triage for a different workflow, and the accounts payable and order processing examples for other process shapes.

What to do next

Write the job-request route, run identical normal and failure cases in both tools, and count use according to each current plan. Assign the person who will maintain credentials and recovery. For a separate review of where automation fits the business, take the process map to AI Foundation Audit.

A plan comparison works only when both tools process the same case and recovery path.

In words, the tree branches like this: define a real workflow, test plan access, measure executions and tasks separately, then decide whether the business can operate self-hosting and recover failed runs before committing.