n8n Pricing (2026): Pro Wins Until You Can Own the Server
n8n costs $0 self-hosted or $24 monthly in Cloud. Compare every plan, annual savings, execution overages, and the self-hosting break-even.
n8n costs $0 if you self-host Community Edition, $24 month-to-month for Cloud Starter, or $60 for Cloud Pro. Pro is the best paid default once production workflows exceed Starter's 2,500 monthly runs; Community is cheaper only when someone truly owns updates, backups, security, and incidents.
n8n pricing at a glance
n8n prices its managed plans by workflow execution, meaning one complete run from trigger to finish, not by each internal step. That makes the sticker price look higher than Make for simple flows and much cheaper than task-based tools when a workflow has many actions.
The US prices below were verified against n8n's live pricing page and checkout data on August 3, 2026. Annual figures are prepaid totals, not monthly contracts with a discount applied later.
The blunt recommendation is simple. Start on Starter only if 2,500 executions and one shared project leave comfortable room. Pro-1 is the safer production default. Pro-2 is the capacity upgrade most price comparisons miss. Business is not the next volume tier: it costs much more than Pro-2, is self-hosted, and includes 10,000 fewer executions. Buy it for SSO, environments, Git-backed changes, and organizational control.
Is n8n free? Community Edition is, Cloud is not
n8n Community Edition is free indefinitely, but it is free software rather than a free hosted service. You install it on infrastructure you control, pay no n8n license fee, and operate the database, encryption keys, backups, updates, capacity, and recovery path yourself. The Community Edition documentation says the registered version is also free and adds folders, Debug in Editor, and custom execution data after email registration.
n8n Cloud has no permanent free plan. Starter and Pro offer trials with no credit card, and the trial ends without charging unless you manually subscribe. The current trial package uses Starter limits: 2,500 executions, 5 concurrent executions, 320 MiB RAM, 10 millicore burstable CPU, and 800 AI Assistant credits. The Business trial runs for 14 days, provides Business features, and requires a card.
Community includes nearly the complete workflow-building core, but the exclusions matter as soon as more than one person owns production. It omits Projects, shared workflows and credentials, SSO, Git version control, external secret stores, external binary storage, log streaming, and multi-main. Queue mode is available, but multi-main, the high-availability control layer, is not. Those are not cosmetic upgrades. They are the controls that stop credentials, deployments, and incident response from depending on one operator's memory.
The license is another boundary. n8n is source-available fair-code, not OSI open source. Its Sustainable Use License permits internal business, personal, and non-commercial use. It does not permit white-labeling n8n and selling access, or hosting n8n as the substantial paid product, without a separate agreement. n8n's own examples also distinguish a product backend using the company's credentials, generally allowed, from collecting an end user's third-party credentials for an n8n-powered feature, which is not allowed under the published example.
Who never needs to pay n8n? A solo technical builder or internal automation team can stay on Community when the license fits, production controls are modest, and one named person owns the infrastructure. The n8n bill can remain $0 even at meaningful execution volume. The server, backups, monitoring, model APIs, and operator time do not become free with it.
One execution is a whole workflow run
Frequency, not complexity, is the first n8n cost driver. One workflow that fires and completes counts as one execution whether it has 2 nodes or 50. A complex lead-enrichment flow can therefore cost the same n8n unit as a one-step Slack alert. That is the economic advantage over platforms that meter each successful action or module.
The catch is that a tiny scheduled workflow can consume a plan quickly. A job firing every five minutes runs 288 times per day, 8,640 times in a 30-day month, or 8,928 times in a 31-day month. That one schedule almost fills Pro-1 and cannot fit Starter. A webhook processing 50 inbound leads a day runs about 1,500 times in a 30-day month and fits Starter. A support bot handling 100 conversations a week with 8 separately processed messages per conversation creates about 3,200 monthly executions over four weeks and needs Pro.
Count scheduled runs
Convert each interval into monthly executions. Daily is about 30 or 31. Every five minutes is 8,640 to 8,928. Count retries and separate schedules as separate runs when they trigger the workflow again.
Count event-driven runs
For forms, webhooks, orders, or CRM changes, use the number of monthly events. Fifty daily events mean roughly 1,500 monthly executions before retries or manual replays.
Count message-driven runs
For chat and agent workflows, multiply conversations by messages that trigger automation. One conversation is not automatically one execution if every turn starts the workflow again.
Choose the first tier with room
Map the result to 2,500, 10,000, or 50,000 managed executions. Do not use the exact cap as the forecast. A plan that only fits a perfect month has no room for retries, growth, or an unusually busy week.
The broader n8n vs Zapier vs Make comparison is useful once the execution count is known. Without that count, comparing plan names is procurement theater.
Three workload budgets that expose the right plan
A plan estimate becomes useful only when it maps a real trigger pattern to the execution meter. These three models show how the same product can be inexpensive for a long workflow and surprisingly costly for a tiny one.
A local-services lead pipeline: Starter with room
Suppose 50 form submissions arrive each day and each submission triggers one workflow. The workflow validates the lead, enriches the company, updates the CRM, assigns an owner, and sends a confirmation. Those five internal actions still consume one n8n execution per lead, so a 30-day month produces about 1,500 executions.
Starter fits and leaves 1,000 of its 2,500 runs unused. That is a reasonable buffer for retries, manual replays, and busier days. The relevant risk is not adding another node inside the same workflow. It is adding a second trigger, such as a separate follow-up workflow that runs for every lead. If that new workflow fires once per lead, monthly use doubles to roughly 3,000 and Pro-1 becomes the honest budget.
A five-minute inventory sync: Pro-1 with little room
A single scheduled workflow running every five minutes creates 8,640 executions in a 30-day month and 8,928 in a 31-day month. Pro-1 technically fits, but only 1,072 to 1,360 executions remain. That is 10.7% to 13.6% of the allowance, before retries, manual tests, or any other production workflow.
This is the case where a tiny flow can cost more than a complicated sales pipeline. Reducing its node count does nothing to the n8n bill. Running it every ten minutes would cut the execution count in half; adding a second five-minute sync would break the Pro-1 allowance. Schedule design matters more than visual complexity.
Pro-1 is acceptable when the five-minute service is the main workload and failures rarely replay. Pro-2 is the more durable budget when several scheduled integrations share the account or the polling interval cannot change.
A message-driven support bot: architecture changes the bill
Take 100 support conversations a week with eight customer messages in each conversation. If every message starts a new workflow, four weeks create about 3,200 n8n executions. Starter cannot hold it, so Pro-1 is the entry plan even before model tokens, vector storage, or help-desk charges are added.
If the architecture can collect a conversation and run one post-conversation workflow, the same 400 conversations create roughly 400 executions instead. That does not mean batching is always the right product decision. Customers may need a response after every message. It shows that the trigger boundary is a pricing decision, not just an engineering detail.
For a purchase forecast, model the normal month, the busiest known month, and a failure month with retries. Then keep 20% to 30% operating room as an editorial rule of thumb, not an n8n requirement. A plan that fits only the average is already too small for production.
What every n8n plan costs and where it stops
Starter: $24 monthly or $240 yearly
Starter is the right managed plan for low-frequency automation, not a production bargain for every small business. It includes 2,500 monthly executions, unlimited users and workflows, one shared project, five concurrent executions, 2,300 monthly AI Assistant credits, and forum support.
At month-to-month pricing, the allowance costs $9.60 per 1,000 included executions. Annual prepayment lowers that to $8.00 and saves $48 over twelve monthly payments. The limit that usually forces the upgrade is execution frequency, followed by the single shared project and five-run concurrency ceiling.
Skip Starter when one five-minute schedule exists, more than one working group needs project separation, or a busy burst can create more than five simultaneous runs. Cheap automation that queues during the event it was meant to handle is not cheap.
Pro-1: $60 monthly or $600 yearly
Pro-1 is the best default for a small production team because it raises the useful limits without handing the team an infrastructure job. It includes 10,000 monthly executions, three shared projects, 20 concurrent executions, admin roles, global variables, workflow history, execution search, 7 days of insights, and 5,700 monthly AI Assistant credits.
The annual equivalent is $50 per month, saving $120 a year. Full utilization costs $6.00 per 1,000 executions month-to-month or $5.00 on annual billing. Moving from Starter to Pro-1 adds 7,500 executions for $30 more at annual equivalents, or $4.00 per additional 1,000.
Pro-1 fits a founder or operations team with several live workflows, light collaboration, and fewer than roughly 10,000 monthly triggers. It stops being comfortable when scheduled work, chat messages, or customer events consistently push near the cap. It also does not provide the SSO and environment controls that turn workflow changes into a governed software process.
Pro-2: $145 monthly or $1,450 yearly
Pro-2 is the hidden value tier for teams that need throughput but not Business governance. n8n's current support documentation names a 50,000-execution Pro-2 tier with 50 concurrent executions and 13,700 monthly AI credits. The live US checkout data prices it at $145 month-to-month or $1,450 prepaid annually, a $120.83 monthly equivalent.
That produces the lowest published managed unit cost in the current ladder: $2.90 per 1,000 executions monthly or about $2.42 annually. Moving from Pro-1 to Pro-2 adds 40,000 executions for $70.83 more per month at annual equivalents, about $1.77 per additional 1,000 executions.
Business: $800 monthly or $8,000 yearly
Business buys organizational controls, not cheaper automation. The base plan is self-hosted, includes 40,000 monthly executions, and adds six shared projects, SSO through SAML and LDAP, separate environments, scaling options, Git version control, 30 days of insights, and forum support.
The annual total is $8,000, a $666.67 monthly equivalent and a $1,600 saving against twelve $800 payments. At full base allowance, that is $20 per 1,000 executions month-to-month or $16.67 annually. Both figures are much higher than Pro-2, and Business supplies 10,000 fewer included runs. The premium pays for control over identity, deployments, and team boundaries while keeping the data plane self-hosted.
Business is reasonable when several teams share credentials, changes must pass through environments and Git, or identity policy requires SSO. It is hard to defend for one founder and an operations lead. Dedicated support with an SLA is also not part of Business; that begins with Enterprise.
Enterprise: custom price for custom controls
Enterprise is the plan for compliance, contractual support, and scale that cannot fit a self-serve card. Pricing and execution volume are custom. It can be hosted by n8n or self-hosted and adds unlimited shared projects, 200-plus concurrent executions, 365 days of insights, external secret storage, log streaming, extended data retention, SLA support, and invoice billing.
Buy Enterprise when those controls are procurement requirements or outage response needs a contracted escalation path. Skip it when the desired feature is merely more executions. Pro-2 or a custom Pro capacity request is the relevant comparison first.
- One full workflow run is billed once, regardless of internal step count.
- Unlimited users, workflows, steps, and integrations remove common expansion fees.
- Pro-2 reaches 50,000 managed executions before governance pricing begins.
- Community removes the n8n license fee for internal use when self-hosting fits.
- The permanent free option is self-hosted, so infrastructure ownership moves to you.
- Public Cloud overage pricing is not disclosed on the current help pages.
- Business is self-hosted, expensive, and includes fewer runs than Pro-2.
- Annual Cloud payment is locked for 12 months and is non-refundable.
The break-even math between n8n tiers
Starter flips to Pro-1 the moment expected monthly runs no longer fit 2,500 or the team needs Pro features. There is no sensible way to value the 2,501st run as an isolated overage because n8n does not publish a Starter overage price. The decision is the full $30 annual-equivalent jump from $20 to $50, buying 7,500 more executions plus projects, concurrency, admin controls, and workflow tooling.
Pro-1 flips to Pro-2 on cleaner economics. An extra $70.83 per month at annual equivalents buys 40,000 additional executions, reducing the marginal cost to about $1.77 per extra 1,000. A workload forecast above 10,000 should price Pro-2 before considering infrastructure changes.
Every unit-cost figure in this article assumes the full included allowance is used. Realized cost is higher when runs sit unused. A Starter customer completing 500 workflows in a monthly-billed month pays an effective $48 per 1,000 executions, not $9.60. At 1,000 completed workflows, the figure is $24 per 1,000. The annual equivalents for those same workloads are $40 and $20 per 1,000.
The same utilization rule applies to Pro. A team buying Pro-1 annually for only 3,000 monthly executions realizes a cost of $16.67 per 1,000, even though the plan's full-utilization figure is $5. Capacity is still valuable when it absorbs peaks, but it should be named as insurance rather than disguised as low unit cost. Compare plans using expected use, peak use, and included headroom side by side.
Business has no throughput break-even against Pro-2. It costs $545.84 more per month at annual equivalents and includes 10,000 fewer executions. Its value appears only when SSO, environments, Git version control, and self-hosted control replace risk or manual process worth more than the difference.
The annual lock is easy to compute. Starter saves $48 a year, Pro-1 saves $120, Pro-2 saves $290, and Business saves $1,600. Annual billing is sensible when the usage band and ownership model are stable. Monthly is worth the premium during migrations, seasonal demand, or any period where the workload might move between Cloud and Community.
What changed in n8n pricing, and what to recheck
The current pricing model is materially different from older n8n comparisons. In August 2025, n8n removed paid-plan limits on active workflows and made users and workflow steps unlimited. The primary meter became completed workflow executions. Advice that treats workflow count, seat count, or node count as the main paid-plan constraint is describing an older model.
The present ladder also contains a capacity step that is easy to miss. Pro-2 provides 50,000 managed executions for $145 monthly or $1,450 annually, while the $800 monthly Business base includes 40,000 and must be self-hosted. That inversion is deliberate: Pro-2 sells managed volume, while Business sells identity, deployment, and governance controls. A comparison that jumps from the visible Pro headline to Business can overstate the cost of managed scale by hundreds of dollars a month.
AI Assistant allowances are another refresh signal. The live plan material now lists 2,300 monthly credits for Starter, 5,700 for Pro-1, and 13,700 for Pro-2. Those are build-time assistant credits and not runtime model tokens. Old screenshots with much smaller allowances are not reliable evidence for the current package.
Recheck four items before approving a purchase order: the monthly checkout price, the annual prepaid total, the selected execution band, and whether the plan is Cloud or self-hosted. Then verify the AI-credit allowance separately from the runtime API budget. This article was checked on August 3, 2026; its date is part of the answer because an undated pricing table is a liability.
Self-hosting is free software, not free operations
DigitalOcean shows why the $0 Community label needs a total-cost model. Its current Basic Droplet price is $12 per month for 2 GiB of memory, one vCPU, 2,000 GiB of transfer, and a 50 GiB SSD. Weekly backups add 20%, taking that simple server-and-backup example to $14.40 per month.
That $14.40 is an illustrative cash floor, not a production architecture recommendation. It excludes monitoring, an external database or replication, secret management, alerting, patching, incident response, and the labor required to prove a restore works. A larger workload can also need more CPU, memory, workers, and storage.
The labor break-even is sharper than the server bill. Value technical time at an explicit $75 per hour assumption. Against Starter's $20 annual equivalent, the modeled Community stack can absorb only 4.48 minutes of maintenance per month before it costs more. Against Pro-1 at $50, it can absorb 28.48 minutes. One security update, failed upgrade, or restore test can consume that allowance.
Self-hosting still wins for the right team. It can remove the n8n execution meter, place data inside a private network, permit custom infrastructure configuration, and keep internal workflows on a platform the team already operates. The decision should be made because control or scale earns the operating responsibility, not because $0 looks smaller than $20.
Hidden costs: overages, credits, APIs, refunds, and annual lock
Business has the only concrete public overage price in the current plan material. An extra bucket of 300,000 executions costs EUR 4,000, or EUR 13.33 per additional 1,000 if the whole bucket is used. Workflows continue after the Business quota is exceeded, but n8n may invoice the overage 45 days later if the buyer does not respond or move to an upgrade path.
Cloud Starter and Pro are less transparent. Current help pages describe a one-time courtesy reset for some trial, Starter, and Pro-1 accounts, a Pro-2 self-service path up to 500,000 monthly executions, and a support request for annual Pro-2 capacity. They do not publish the current dollar price for those extra Cloud executions. Treat a forecast above 50,000 as a quote request, not as a linearly extrapolated bargain.
The reset date is another budgeting wrinkle. Cloud execution and AI-credit allowances reset on the first of each month, independent of the subscription billing date. A customer starting mid-month should not assume one full allowance lasts until the next card charge.
AI Assistant credits are for help building workflows, not a runtime model budget. Starter includes 2,300 per month, Pro-1 5,700, and Pro-2 13,700. Unused credits do not roll over and cannot currently be purchased separately. Self-hosted AI Assistant is listed as coming soon and will use the customer's own API key.
OpenAI, Anthropic, Google, database, messaging, and other third-party bills remain separate. n8n's self-serve terms explicitly separate third-party subscription fees from the n8n subscription. A $50 Pro plan does not include the tokens consumed by an AI agent at runtime. The workflow platform bill and the model bill need separate alerts.
Refunds split by commitment. The current n8n refund policy gives monthly customers a refund for that month's payment when requested within 14 days of the billing date and the account is cancelled. Annual Cloud payments are prepaid and non-refundable under the current self-serve terms. Listed prices also exclude applicable taxes and duties.
The free trial is unusually low risk for Cloud Starter and Pro because no card is required and the account is not charged automatically. Export workflows before leaving. The current help page says trial workflow data is retained for 90 days, while access ends after the trial unless a paid plan is chosen.
Discounts are narrow. An individual student cannot redeem n8n's education discount. Educational institutions can request case-by-case pricing. Eligible startups can receive 50% off Business when they have fewer than 20 employees and less than EUR 5 million in funding, but the discount does not remove the self-hosting requirement.
n8n vs Make vs Zapier pricing, normalized
Make has the lowest entry paid sticker for simple automation. Its current Core plan costs $12 per month for 10,000 credits; Free includes 1,000 credits and a 15-minute minimum schedule interval. Most non-AI module operations consume one credit, while some AI and advanced modules use dynamic credit rates.
For a five-action outcome, Make Core's 10,000 credits support 2,000 outcomes if every action consumes one fixed credit. At full use, that is $0.006 per outcome. It beats n8n Starter's $0.0096 month-to-month unit cost on this simple model and carries a lower minimum bill.
Zapier starts Professional at $19.99 per month, with the public selector beginning at 750 tasks. A task is normally one successful action; triggers, filters, paths, errors, and some built-in steps do not count. Zapier's strength is the convenience of its app ecosystem and a familiar hosted operating model, but multi-action Zaps consume the allowance quickly.
A five-action outcome consumes five ordinary Zapier tasks. The starting 750-task allowance therefore supports 150 such outcomes. At the displayed $19.99 starting price, that is about $0.133 per outcome at full use. Higher task tiers have different unit rates, so this is an entry-tier normalization rather than a forecast for every Zapier account.
n8n Starter supports 2,500 five-action outcomes because the five steps still count as one execution. Its $24 monthly price is $0.0096 per outcome at full use. The result is not that n8n always wins. Make Core is cheaper in this entry scenario. The result is that n8n's economics improve as step count rises, while Make and Zapier consume more billable units for the same completed workflow.
- Choose Make when workflows are simple, the $12 minimum matters, and credit consumption is easy to predict.
- Choose Zapier when connector convenience and faster non-technical setup earn more than the task premium.
- Choose n8n Cloud when workflows are step-heavy, code and HTTP flexibility matter, and managed hosting is still worth paying for.
- Choose n8n Community when infrastructure ownership and the license boundary are already solved.
Teams that reject one of those constraints can compare the best n8n alternatives. The broader AI automation tools shortlist covers different orchestration and agent-building shapes.
Which n8n plan should you choose?
Choose Community when the workflow is for internal business use, a technical owner already operates comparable services, and the missing governance features do not matter. Do not choose it merely to save $20. A few minutes of monthly labor can erase that difference.
Choose Starter when expected runs stay comfortably below 2,500, one shared project is enough, and bursts will not exceed five concurrent executions. It is a proving-ground plan for light production, not the safe default for a growing event stream.
Choose Pro-1 when production reliability matters, run volume fits below 10,000, and the team wants managed hosting with more projects and concurrency. This is the default small-business choice.
Choose Pro-2 when execution volume, not governance, is the problem. Its 50,000 managed runs and lower unit price make it the rational capacity step before self-hosted Business.
Choose Business when SSO, separate environments, Git version control, shared project boundaries, and self-hosted control are active requirements. Choose Enterprise when dedicated support, external secrets, log streaming, extended retention, or a custom hosted contract is mandatory.
The decision flips on ownership first, executions second, and governance third. If nobody owns infrastructure, stay in Cloud. If managed runs exceed 10,000, price Pro-2. If identity and deployment controls are mandatory, price Business against Enterprise rather than against Starter.
n8n pricing FAQ
Is it free to use n8n?
Yes. n8n Community Edition has a $0 license fee and can be self-hosted indefinitely for permitted internal, personal, or non-commercial use. n8n Cloud has a no-card trial but no permanent free plan, so a subscription is required after the trial.
How much is n8n per month?
US month-to-month pricing is $24 for Starter with 2,500 executions, $60 for Pro-1 with 10,000, $145 for Pro-2 with 50,000, and $800 for self-hosted Business with 40,000. Annual prepayment lowers the equivalents to $20, $50, $120.83, and $666.67. Enterprise is custom.
What happens when n8n exceeds its execution limit?
The route depends on the plan. Starter and Pro customers move to a higher execution tier or arrange custom capacity; current help pages do not publish a universal Cloud overage price. Business can continue running and may be billed EUR 4,000 for an extra 300,000-execution bucket.
Is n8n cheaper than Zapier?
n8n is usually cheaper for step-heavy workflows because one full run consumes one execution, while Zapier normally counts each successful action as a task. Zapier can still be the better purchase when its connector coverage and easier setup save more operator time than the task premium.
Does n8n have a student discount?
n8n does not offer an education discount that an individual student can redeem. Educational institutions can apply for case-by-case pricing, and individual users can self-host Community Edition without an n8n license fee.
Can I get an n8n refund?
Monthly customers can request a refund for that month's payment within 14 days of the billing date if they cancel the account. Annual Cloud payments are prepaid for 12 months and non-refundable under the current self-serve terms.
Did n8n pricing change?
Yes. In August 2025 n8n removed active-workflow limits from paid plans and made users and steps unlimited, shifting the main price meter to completed workflow executions. The live prices and allowances above were re-verified on August 3, 2026.
Get the AI Business Workflow Audit Checklist and the operator newsletter.
Aug 3, 2026
