Quick Answer
n8n pricing runs €24 per month for 2,500 workflow executions (Starter), €60 for 10,000 (Pro), and €800 for 40,000 (Business), with 17 percent off on annual billing and custom Enterprise pricing above that. Every plan includes unlimited users and unlimited workflows. The self-hosted Community Edition is free with unlimited executions; you pay only for your server.
An employee at Delivery Hero gets locked out of an account, and the recovery grind begins: a ticket, a queue, an average of 35 minutes gone. Across a global food delivery company, lockouts alone were burning serious time every month.
So the IT team built one n8n workflow. An employee requests recovery, their manager approves, and the workflow resets the account through the connected systems automatically.
Average lockout time fell from 35 minutes to 20, which removed 200 hours of employee downtime per month, according to n8n’s published case study. Dennis Zahrt, Delivery Hero’s director of global IT service delivery, put it plainly: “We have seen drastic efficiency improvements since we started using n8n.”
n8n, pronounced “n-eight-n,” is an open source workflow automation platform: it connects a company’s apps, data, and AI models, then runs the connective work automatically. Teams build workflows in a visual editor, host them on n8n’s cloud or their own servers, and pay based on how often those workflows run.
One workflow, 200 hours back. Numbers like that explain why n8n’s repository carries more than 200,000 stars on GitHub, and why its bill lands on more finance desks every quarter. What lands with it is a pricing model unlike the tools it replaces, in euros, with a usage meter AI agents can spin faster than anyone budgets for.
How much does n8n cost in 2026?
Paid cloud plans start at €24 a month and the self-hosted Business tier tops the published lineup at €800, with Enterprise negotiated above it. The variable that actually moves your bill is workflow executions, because users and workflows are now unlimited on every plan.
| Plan | Monthly billing | Annual billing (per mo) | Executions/mo | Concurrent runs | Hosting |
|---|---|---|---|---|---|
| Community Edition | Free | Free | Unlimited | Self-managed | Self-hosted |
| Starter | €24 | €20 | 2,500 | 5 | n8n Cloud |
| Pro | €60 | €50 | 10,000 | 20 | n8n Cloud |
| Business | €800 | €667 | 40,000 | 30 | Self-hosted |
| Enterprise | Custom | Custom | Custom | 200+ | Either |
Concurrent executions caps how many workflow runs can execute at the same instant; hit the cap and new runs queue rather than fail. It governs speed, not your monthly execution total.
Prices are in euros; n8n is headquartered in Berlin and hosts cloud data in Frankfurt. Annual-billing figures appear directly on n8n’s pricing page, and monthly figures follow its published 17% annual discount.
Plans differ on more than volume: Starter includes one shared project; a shared project is n8n’s container for grouping workflows and credentials so a team can collaborate under one set of access permissions. Starter also includes 2,300 AI Assistant credits, an AI Assistant credit being n8n’s usage unit for its built-in AI Assistant features, separate from workflow executions.
Pro adds admin roles, workflow history, and up to 13,700 AI credits, Business adds SSO, Git version control, and environments, and Enterprise adds log streaming, a dedicated support SLA, and invoice billing.
Report
Finance needs to prove AI’s return: CloudZero report
260 senior finance leaders (more than half CFOs) told us why the speed of seeing AI spend, not the size of it, separates who pulls ahead on AI from who gets burned.
How does execution-based pricing actually work?
One execution is a single run of your entire workflow, no matter how many steps it contains or how much data it touches. A two-step workflow and a fifty-step workflow each consume exactly one execution per run. Per-task competitors meter every action inside the run, which is why the same automation can cost dramatically different amounts on different platforms.
That makes the per-run price easy to compute, and cheap rates come with an expensive catch: unused executions don’t roll over. A team consuming 1,200 of Starter’s 2,500 executions pays an effective €0.020 per run, double the list rate, month after month.
CloudZero calls that gap the difference between a rate and a unit cost, the same discipline that governs AI agent pricing models across the market. Size the plan to measured run volume and recheck quarterly.
What underuse does to the real per-run price, in CloudZero’s arithmetic on monthly list rates:
| Plan | List price per execution | Effective price at 50% usage |
|---|---|---|
| Starter (€24, 2,500 runs) | €0.010 | €0.019 |
| Pro (€60, 10,000 runs) | €0.006 | €0.012 |
| Business (€800, 40,000 runs) | €0.020 | €0.040 |
Is n8n free?
Yes, genuinely, in one form: the self-hosted Community Edition is free software with unlimited executions and unlimited workflows. The trade is that you run it yourself, so the true cost is a server plus the operations time to patch, back up, and scale it.
The cloud plans answer the free-tier question differently: Starter and Pro offer trials with no credit card and 800 AI Assistant credits, while the Business trial requires a card. After the trial, the meter starts.
Which route wins depends on volume and staffing. High-volume teams with an engineer to spare often self-host and pocket the difference. Teams that would rather ship workflows than babysit infrastructure pay the cloud premium and get managed uptime for it.
What does self-hosted n8n really cost?
An n8n self-hosted pricing estimate has three lines: the free software, the server underneath it, and the licensing quirks above it. A small VPS runs a few euros a month; production-grade hosting with backups and monitoring costs more, and it belongs in your cloud cost management picture like any other compute.
The quirk most buyers miss: Business, the €800 tier, is itself self-hosted. You pay for features and license, not hosting, and the license key phones n8n’s license server daily with production execution counts to stay active. Paid self-hosting means metered self-hosting.
The Community Edition carries no such meter. What it lacks is the collaboration, SSO, and governance layer that makes automation survivable at company scale, which is precisely the layer finance and security teams eventually require.
What do AI agents do to your execution count?
They multiply it, and faster than most budgets expect. n8n’s own estimation guidance makes the arithmetic vivid: a single workflow triggered every five minutes runs roughly 8,600 to 8,900 executions a month. One scheduled workflow nearly exhausts an entire Pro plan by itself.
Agent and chatbot workloads scale the same way. Using n8n’s method of estimating conversations times messages, a support bot handling 500 conversations a week at eight messages each generates about 17,000 executions a month, past Pro’s ceiling before any other automation runs. Executions are the tokens of automation: a metered unit that climbs with adoption, exactly like the token spend behind the models themselves.
And the n8n line rarely travels alone. Agent workflows chain executions to model API calls, so every run can pull OpenAI or Anthropic billing along with it. Modeling the workload before committing, the way CloudZero approaches LLM cost estimation, keeps the AI agent cost picture whole instead of scattered across invoices. One n8n user, Ollie Scheers, described the platform’s role this way: “Tools like ChatGPT and Claude are great, but n8n is the thing that allows you to integrate AI into your work and your processes in a safe and controlled way.”
That’s the shift worth budgeting for. Wiring agents into n8n converts a flat automation subscription into usage-based agentic AI spend, with all the forecasting problems that phrase implies.
What happens when you outgrow your plan?
On Starter and Pro, workflows pause when you hit your execution limit, which protects your wallet and interrupts your operations in the same moment. On Business and Enterprise, workflows keep running and overages apply: n8n bills €4,000 for each extra bucket of 300,000 executions, invoiced automatically 45 days after they occur if you haven’t upgraded.
The arithmetic there deserves a second look. That overage bucket prices out to roughly €0.013 per execution, below the €0.020 Business list rate, so a moderate overage isn’t the disaster the word suggests. The real risk is the surprise: €4,000 landing six weeks after the usage that caused it, on a bill nobody connected to the agent launch that drove the volume.
The fix is boring and effective: alert on execution consumption the way you alert on any usage-based line, treat those alerts as routine AI cost optimization work, and fold plan reviews into the same AI budgeting cadence that governs your model spend.
What does n8n Enterprise cost?
There’s no published number for n8n enterprise pricing: it’s negotiated, sized on execution volume, and sold with 200-plus concurrent executions, 365 days of insights, external secret store integration, log streaming, dedicated support under SLA, and invoice billing. Hosting is your choice, n8n’s cloud or your own infrastructure.
Two levers matter in that negotiation. Companies under 20 employees with less than €5 million in total funding can skip it entirely for one year: n8n’s published Startup Plan takes 50 percent off Business. Everyone else should walk in knowing their real monthly run volume, because a custom execution quota priced against guessed usage repeats the pay-for-peak mistake with more zeros attached.
How does n8n pricing compare with Zapier and Make?
Most buyers frame this as n8n vs. Zapier math, and it comes down to what gets metered. Zapier bills per task, though triggers and several built-in steps (Filter, Formatter, Paths, Delay) run free. Make now bills in credits, the term that replaced “operations,” with AI modules metered dynamically on tokens rather than a flat per-step rate. n8n bills per run: a ten-step workflow executed 1,000 times consumes 1,000 executions on n8n and 10,000 tasks on a per-task platform.
Complexity decides the winner. Simple two-step zaps at low volume can cost less on per-task tools with cheaper entry points, while long, branching workflows favor n8n’s model by widening margins with every added step. The honest comparison prices your actual workflows on each meter, the same way any AI pricing evaluation should run on your workload rather than the vendor’s example.
The deeper plan-by-plan comparison deserves its own treatment, and dollar figures for the per-task platforms belong there, verified on their pages rather than rounded from memory here.
How do you keep automation spend accountable?
Treat n8n as one meter in a larger machine. The workflow that saves 200 hours also burns executions, model tokens, and cloud compute, and the finance question is never what the tool costs but what the outcome costs. That’s a unit economics question: cost per recovered account, per processed order, per resolved ticket.
Answering it means allocating automation spend the way disciplined teams already allocate cloud spend: by team, product, and workload, so a spike traces to the agent that caused it within hours instead of a billing cycle. CloudZero’s work on AI spend keeps finding the same pattern, and its AI cost management framework exists because the pattern generalizes: usage-based tools reward the companies that measure them and quietly tax the ones that don’t.
Get that measurement in place and the AI ROI conversation changes shape. Delivery Hero could name its return because it measured the hours saved. That’s the demand side of ROI. The cost side, what the automation itself runs, is the one most programs never quantify, and it’s the one the CFO signs.
Leading global organizations such as Toyota, Skyscanner, Grammarly, Duolingo, Upstart, and Coinbase use CloudZero to see exactly where their cloud and AI spend goes, automation included. Request a demo to see your AI spend allocated by team and product, get a free cloud cost assessment, or take the self-guided product tour.
n8n pricing FAQs
