For most small businesses, n8n Cloud Starter costs €24 month-to-month or €20 per month billed annually, while Pro costs €60 month-to-month or €50 per month billed annually. The default allowances shown on n8n’s pricing page are 2,500 workflow executions for Starter and 10,000 for Pro. Business is €800 month-to-month or €667 per month billed annually with 40,000 executions, but it is a self-hosted commercial plan rather than a managed Cloud upgrade.
The crucial detail is what n8n counts: one complete workflow run is one execution, even if it contains many steps. That can make complex workflows cheaper than products that meter every action, but a high-frequency trigger can burn through an allowance quickly. Self-hosting the free Community Edition can reduce software fees, but only if your business already has someone who can reliably own hosting, backups, upgrades, monitoring, and security.
Pricing and limits were checked against n8n’s first-party pricing and documentation on August 17, 2026. The pricing page localized to euros in our verification region; taxes, currency, checkout pricing, and negotiated Enterprise terms can differ. This is a vendor-document review, not hands-on product testing.
Managed Cloud converts infrastructure work into a subscription; self-hosting converts it into operational responsibility. Editorial illustration, not a product screenshot.
What does n8n cost in 2026?
The practical SMB range starts at €20 per month on an annual Starter commitment and reaches €667 per month for the annual Business plan before infrastructure and labour. Enterprise pricing is custom.
| Plan | Hosting | Monthly billing | Annual billing rate | Default execution allowance | Published fit |
|---|---|---|---|---|---|
| Starter | n8n Cloud | €24/month | €20/month (€240/year) | 2,500/month | Getting started and lighter production use |
| Pro | n8n Cloud | €60/month | €50/month (€600/year) | 10,000/month | Solo builders and small production teams |
| Business | Self-hosted commercial | €800/month | €667/month (€8,004/year) | 40,000/month | Companies under 100 employees needing collaboration and scale |
| Enterprise | Cloud or self-hosted | Custom | Custom | Custom | Governance, compliance, scale, and support requirements |
| Community Edition | Self-hosted | €0 software subscription | €0 software subscription | Not metered as a paid Cloud allowance | Technical teams willing to operate the instance |
Annual billing is presented as a 17% saving. Treat the figures above as the public euro price card, not a universal quote: verify the currency, tax treatment, execution tier, and billing period in n8n’s checkout before approval.
Business deserves special attention. It adds six shared projects, SSO through SAML or LDAP, separate environments, Git version control, and scaling options, but your company still operates the hosting. Enterprise adds custom executions, unlimited shared projects, external secret-store integration, log streaming, extended retention, an SLA, and invoice billing; the final limits and price are contractual.
What is included in each n8n plan?
All paid plans advertise unlimited users, unlimited workflows, and all integrations; the meaningful limits are executions, concurrency, saved history, collaboration controls, and governance.
| Published limit or entitlement | Starter | Pro | Business | Enterprise |
|---|---|---|---|---|
| Shared projects | 1 | 3 | 6 | Unlimited |
| Concurrent executions | 5 | 20 | Not numerically stated on the public card | 200+ |
| Workflow-history maximum | 1 day | 5 days | Git/version-control features included | 365+ days |
| Execution-log retention | 7 days | 30 days | Public card does not state a numeric limit | Extended/custom; comparison table shows unlimited |
| Saved executions | 2,500 | 25,000 | Not publicly specified on the card | 50,000 shown in Cloud comparison |
| Saved-execution storage | 2.5 GB | 25 GB | Operated in your own environment | 50 GB shown in Cloud comparison |
| Insights window | Basic | 7 days | 30 days | 365 days |
| Admin roles | No dedicated admin-role feature listed | Included | Included via Pro feature inheritance | Included |
| SSO (SAML/LDAP) | No | No | Included | Included |
| Support | Forum | Forum, with email support shown in comparison | Forum; no dedicated support | Dedicated support with SLA |
Do not confuse workflow history, execution-log retention, saved-execution count, storage, and Insights. They are different controls. n8n says old execution records are removed when the first applicable saved-count, storage, or retention limit is reached; workflows continue running, but less history remains available for debugging.
The public page does not state every Business-plan capacity as a clean number. Ask sales or verify the licence terms if concurrency, retention, queue mode, worker capacity, or environment design will influence your purchase.
How n8n execution billing differs from tasks and operations
n8n charges paid plans by complete production workflow execution, not by every step inside the workflow. A 15-step workflow that runs once normally consumes one n8n execution. In a step-metered system, the same run may consume multiple tasks, operations, or credits.
n8n’s execution unit can favour long workflows; step-based products multiply usage as actions and bundles move through the scenario.
This pricing shape matters more than the sticker price:
- A nightly 20-step finance sync uses roughly 30 n8n executions per month, not 600 billable steps.
- A webhook that fires 6,000 times per month uses about 6,000 executions even if each run has only two steps.
- A workflow scheduled every five minutes runs roughly 8,600 to 8,900 times per month, before retries or separate sub-workflows.
- Breaking one process into several called workflows can increase execution consumption, so architecture still affects cost.
Make now describes credits as its billing currency. For typical non-AI modules, one module operation equals one credit, and bundles can cause downstream modules to run repeatedly. Zapier’s public pricing explains that each successful action generally consumes a task, with AI steps, code, connectors, MCP, and related products drawing from a shared task pool at rates that may vary. In plain English: n8n meters the run; Make and Zapier more often meter work inside the run.
Are n8n AI Assistant credits separate from model API charges?
Yes. AI Assistant credits pay for n8n’s in-product workflow-building assistant; they are not a prepaid balance for OpenAI, Anthropic, Google, or another model used inside your automated workflows.
Starter includes 2,300 AI Assistant credits per month. Pro includes 5,700 or up to 13,700 depending on the selected plan size; the default 10,000-execution card displays up to 13,700. Credits refresh monthly, do not roll over, and n8n says additional credits cannot currently be purchased separately. Trial accounts receive 800 credits that expire when the trial ends.
If an automation calls an external model through your API key, the model provider bills that usage separately. Those API charges sit on top of n8n hosting or licence costs. Self-hosted AI Assistant is marked “coming soon” on the pricing page; n8n says it will use your own API key, so there would be no n8n Assistant credit allowance, but model-provider charges would still apply.
What will low-, moderate-, and high-frequency workflows cost?
Workflow frequency usually determines the Cloud plan before user count does, because users and workflows are advertised as unlimited. These scenarios use the public default allowances and annual billing rates.
| SMB scenario | Estimated monthly executions | Likely public plan | Monthly subscription | Annual subscription | Main risk |
|---|---|---|---|---|---|
| Low: daily reporting, weekly backups, 25 lead handoffs/day | 900 | Starter | €20 | €240 | A new webhook process can unexpectedly consume the remaining headroom |
| Moderate: 200 orders/day plus scheduled finance and alerts | 7,000 | Pro | €50 | €600 | Retries, chat messages, or split workflows can push use past 10,000 |
| High: 1,000 customer events/day | 30,000 | Business default allowance or a custom Cloud quote | €667 plus hosting for Business | €8,004 plus hosting | Business is self-hosted; a managed Cloud requirement changes the quote |
The estimate should start with triggers, not steps. Count scheduled runs, webhook events, chat messages, polling schedules, error workflows, and deliberately separate sub-workflows. Add 20% to 30% headroom for growth and retry behaviour, then compare the result with the allowance actually shown at checkout.
The high-frequency example does not prove Business is the right answer. An intermediate Pro execution tier may appear in the live selector, and Enterprise Cloud may be preferable when the company does not want to run infrastructure. Get a written quote once the default card no longer matches the workload.
What is the real cost of self-hosting n8n?
Self-hosting is not “free n8n”; it is a transfer of cost from the vendor invoice to infrastructure and staff time. Community Edition removes the software subscription, while Business and Enterprise self-hosting still require paid licences.
The server bill is only one line: backups, upgrades, monitoring, security, and operator time complete the self-hosting cost.
The following planning scenarios are illustrative assumptions, not vendor quotes. Labour is valued at €50 per hour and excludes initial migration or workflow-building time because both Cloud and self-hosted deployments require that work.
| Self-hosted scenario | Infrastructure | Backups/monitoring | Ongoing staff time | Estimated monthly TCO | Estimated annual TCO |
|---|---|---|---|---|---|
| Low: single small instance, non-critical workflows | €10 | €5 | 1 hour (€50) | €65 | €780 |
| Moderate: production instance with tested backups and alerts | €25 | €10 | 2 hours (€100) | €135 | €1,620 |
| High: redundant or larger deployment with stricter operations | €80 | €20 | 5 hours (€250) | €350 | €4,200 |
These estimates can be too low when the business needs high availability, separate staging, managed databases, object storage, log aggregation, incident response, penetration testing, or formal compliance evidence. They can also be too high when a competent operator already maintains the same platform and the incremental workload is genuinely small.
When is self-hosting genuinely cheaper?
Self-hosting is genuinely cheaper when execution volume is high, Community Edition has the features you need, and infrastructure work fits into an existing technical role. It is usually a false economy when the owner, agency, or senior developer must repeatedly rescue a neglected server.
Choose Community Edition self-hosting when:
- Pro’s execution allowance is too small but Business governance features are unnecessary.
- The team already operates containers, databases, TLS, backups, secrets, monitoring, and patching.
- Downtime has a defined response owner and recovery procedure.
- Data-location or network-access requirements make your own environment preferable.
- The expected labour cost remains below the Cloud premium after adding realistic incident time.
Choose n8n Cloud when:
- Nobody owns infrastructure operations.
- Starter or Pro covers expected executions with sensible headroom.
- Fast setup and vendor-managed upgrades matter more than maximum cost control.
- The workflows are important enough that “someone checks the server occasionally” is not an operating model.
Choose paid Business or Enterprise when collaboration, SSO, environments, Git, external secrets, log streaming, SLA support, or formal procurement matters. Self-hosting Community Edition to avoid a licence while rebuilding those controls manually is rarely clever; it is just enterprise software wearing a fake moustache.
How does n8n’s cost shape compare with Make and Zapier Agents?
n8n tends to favour long, complex workflows; Make and Zapier become more sensitive to the number and type of actions performed inside each run. This is a cost-shape comparison, not a claim that one product is always cheaper.
| Platform | Primary usage shape | What multiplies usage | Separate AI/model cost risk |
|---|---|---|---|
| n8n | Complete workflow executions | Trigger frequency, retries, and separately executed workflows | Connected model APIs are billed separately; Assistant credits cover the builder assistant |
| Make | Credits, commonly tied to module operations; dynamic AI features can also reflect tokens and other factors | Modules, bundles, polling, tokens, file size, pages, or runtime depending on the feature | Custom AI connections can create both Make-credit use and direct provider token charges |
| Zapier / Zapier Agents | Shared task pool with rates that can vary by AI model tier, code runtime, and connector type | Successful actions and some connector/AI/code usage | Included task consumption and, depending on the connection, external model/service charges |
A six-action lead workflow running 1,000 times is roughly 1,000 n8n executions, while a conventional action-metered design may approach several thousand billable units. But a one-step webhook firing 20,000 times is still 20,000 n8n executions. Model the actual workflow graph and event volume in every platform before comparing subscription cards.
Hidden-cost checklist before choosing a plan
The cheapest plan is the one that covers the whole operating model, not the smallest subscription number. Check each item before signing:
- Execution headroom for retries, chat messages, polling, and seasonal peaks
- Workflow splitting and sub-workflow behaviour
- Saved-execution count, storage, and retention needed for debugging
- Concurrent execution limits during bursts
- External AI model, database, email, SMS, and enrichment API charges
- Cloud tax, currency conversion, and annual-commitment terms
- Self-hosted compute, database, storage, bandwidth, and TLS
- Automated backups plus a tested restore process
- Upgrade testing, maintenance windows, and rollback procedures
- Monitoring, log retention, alerts, and on-call ownership
- Secret management, access control, vulnerability patching, and audit needs
- Staff or agency time, including incident response
- SSO, environments, Git, project roles, SLA, and procurement requirements
Which n8n option fits your small business?
Starter is the sensible default for light automation, Pro is the practical Cloud choice for production SMB use, and Community Edition is the value choice only for technically capable operators.
- Buy Starter if you expect under about 2,000 monthly runs, can tolerate forum-led support, and want the lowest-maintenance entry point.
- Buy Pro if workflows are in production, you need admin roles and execution search, or volume is moving toward 10,000 monthly runs.
- Self-host Community Edition if you have reliable infrastructure ownership and high execution volume but do not need paid governance features.
- Evaluate Business if you specifically need SSO, environments, Git, shared-project capacity, and self-hosted scaling. Add hosting and labour to the €667 annual-billing rate.
- Talk to Enterprise sales if you need managed Cloud above the public Pro fit, custom executions, 200+ concurrency, log streaming, external secrets, extended retention, invoice billing, or an SLA.
Before committing, run a two-week workload estimate or trial, record production executions by trigger, and price the same graph in alternatives. Do not choose from the plan name alone; “Pro” is not a unit of capacity, despite the industry’s heroic effort to make it sound like one.
Frequently asked questions
Is n8n free for small businesses?
Community Edition has no n8n software subscription, but the business pays for hosting and operations. n8n Cloud is paid after the trial.
How long is the n8n free trial?
n8n’s pricing FAQ states that the Business trial lasts 14 days. Starter and Pro trials provide Pro features under Starter limits, include 800 AI Assistant credits, and require no credit card; Business requires a card. Confirm the current trial end date during signup because the public FAQ does not present a separate duration in the Starter/Pro sentence.
Does every node count as an n8n execution?
No. One full workflow run is one execution regardless of the number of steps. Trigger frequency and separately run workflows matter more than node count.
What happens when saved-execution limits are reached?
n8n removes older history when the first applicable count, storage, or time limit is reached. It says this does not stop workflows from running, but it reduces the history available for inspection.
Are n8n AI credits the same as OpenAI credits?
No. n8n AI Assistant credits cover the in-product building assistant. Calls to an external model inside a workflow are charged by that model provider under your connected account.
Does n8n charge per user?
The public pricing page says all plans include unlimited users and workflows. Collaboration features such as shared projects, admin roles, SSO, and project roles still vary by plan.
Is n8n self-hosting cheaper than Pro?
Not automatically. A small server may cost less than Pro, but one or two hours of monthly technical labour can make Cloud cheaper. Self-hosting wins when operations are already covered and execution volume is high enough to justify it.
Are these prices available in every region?
No universal currency should be assumed. The first-party page displayed euros during this verification. Check n8n’s live pricing page and checkout for your region, taxes, plan size, and billing period.
Sources and verification
This guide used current first-party pages for product facts and public pricing, plus first-party Make and Zapier documentation for billing-unit comparisons. It did not use Search Console or GA4 evidence because none was supplied, and it makes no traffic, ranking, customer-outcome, or hands-on-testing claim.