
Multi-tenancy realities when small agencies bundle n8n into their service stack — what works in 2026
A client asks if you can host their n8n too. You already run their site and their campaigns, and saying yes feels natural — until you realise that hosting other people's automation is a different business with its own 2am pages.
This is the wall automation agencies hit somewhere past their fifth client account: n8n on a shared database, workflows tagged by client, and a growing sense that one bad change could spill across everyone. n8n has crossed 190,000 GitHub stars and ships a new version most weeks, but it was not built to be multi-tenant. There is no tenant layer that isolates one client from another inside a single instance.
This piece covers what actually works when a small agency bundles managed n8n into its stack in 2026: the three ways to run it for many clients, the one line in n8n's license you cannot skip, the margin math at 20 clients, and where the model stops making sense. The numbers favour the boring option.
TL;DR
n8n has no native multi-tenancy. The clean, license-safe model is one isolated instance per client, not one shared instance with client tags.
n8n's Sustainable Use License permits consulting and self-hosting, but forbids selling access to n8n itself or white-labeling it for money.
At DANIAN's flat €9 per app, 20 client instances cost €180/month wholesale; retailing at €40–60 each leaves €620–1,020/month in margin.
n8n bills one execution per workflow run; Zapier bills one task per action, so multi-step automations get cheaper to run after the switch.
Heavy clients, past roughly 10,000 executions a day, need more resources. DANIAN adds capacity at +€9 per vCPU, on your approval only.
The agency math
The model only works if the margin is real. Twenty client instances on DANIAN cost €180 a month at €9 each. Sell those at €40–60 per client and you keep €620–1,020 a month, before any setup fees. That is recurring revenue on infrastructure you neither own nor maintain.
Start with one client. You charge a flat monthly fee, you run their automation, and you keep the difference between what they pay and what the instance costs you. The whole model is that gap, repeated.
On DANIAN the instance costs €9 a month, flat. There is no per-execution meter and no per-seat charge, so the cost does not move when a client's workflows get busier within normal limits. That predictability is what makes the retail price safe to quote.
Twenty clients at €9 is €180 a month in wholesale cost. Price each client at €40–60 and the spread is €620–1,020 a month, recurring, on top of your existing retainers. Setup fees sit above that.
For comparison, matching that fan-out with n8n's own Cloud Pro plan at €50 per client would cost €1,000 a month in platform fees alone, before any margin. The flat managed model is built for this pattern.
You can see managed n8n at €9/month and the rest of the catalogue priced the same way.
What white-label actually requires
White-label means three things to a client: your domain, your branding in the login, and your invoice. n8n's own editor keeps n8n branding unless you hold its paid Embed or OEM license. So full white-label of the n8n interface is a licensed, paid path — not something to improvise on a shared instance.
Clients picture white-label as their brand on everything. In practice it is three separate things, and they are not equally easy.
A custom domain is straightforward: each client instance can answer on a subdomain you choose. Branded invoices are yours to send through whatever you already bill with, such as Stripe.
The branded login is where the line sits. n8n keeps its own branding in the editor unless you hold its paid Embed or OEM license, which is a separate commercial agreement. You cannot quietly strip it on a self-hosted instance and stay inside the license.
An honest gap: DANIAN does not ship a white-label control plane today. There is no single dashboard that lets you spin up and rebrand client instances yourself yet. The working setup is per-client instances on your domains, operated for the client, with your invoicing on top — and the n8n editor left as it is unless you license Embed.
The license line you can't skip
n8n is free to self-host, but its Sustainable Use License draws one bright line. You may run n8n for your own business, including building automations you deliver to clients. You may not sell access to n8n itself or white-label it for money. That single distinction shapes the whole reseller model.
n8n is fair-code, not fully open source. Since March 2022 it has shipped under the Sustainable Use License, which restricts use to your own “internal business purposes” and to free, non-commercial sharing.
Inside that frame, n8n is generous about agency work. Building workflows for clients, customising n8n, and setting it up and maintaining it are all named as permitted uses. Charging for that consulting is fine.
Two things are named as not allowed: hosting n8n and charging people for access to it, and white-labeling n8n and offering it to customers for money. Those are the lines a reseller has to stay behind.
The clean reading is simple. Give each client their own instance, and bill for the managed service and your work — not for access to n8n as a product. Keep clients out of the editor where you can, so you are delivering automation rather than reselling the tool.
If you do slip over the line, the license gives a 30-day window to fix it once you are notified, after which the rights are reinstated. Treat that as a safety net, not a plan.
This is operational guidance on how the license reads, not legal advice. For an edge case, n8n's licensing contact is the right place to confirm.
Three ways to run n8n for many clients
There are three realistic setups, and they trade isolation against effort. A shared instance with client tags is cheapest and weakest. One n8n container per client is clean but ops-heavy. A managed instance per client hands the operations to someone else. For most small agencies, the third wins.
| Approach | Isolation | License posture | Who runs ops | Cost at 20 clients |
|---|---|---|---|---|
| Shared instance, client tags | Logical only | Grey | You | One server you maintain |
| One container per client | Full (own DB, credentials, webhooks) | Clean | You | 20 servers you maintain |
| Managed instance per client (DANIAN) | Full | Clean | DANIAN | €180/month wholesale |
Option 1 — One shared instance, tagged by client
This is where most agencies start, and where the shared-database wall appears. Every client's workflows live in one process, one database, and one credential store, separated only by tags or folders. It is the cheapest to run and the weakest to trust. One wrong setting can expose one client's data to another, and the license posture is grey because the separation is so thin. You still own every patch, backup, and uptime alert.
Option 2 — One n8n container per client
Give each client their own n8n container and the isolation problem mostly disappears. Each instance has its own database, its own credentials, and its own webhook URLs, so nothing is shared by accident. The license posture is clean. The cost is operational: 20 clients means 20 containers to patch, 20 backup routines to verify, and 20 sets of webhooks to keep alive. The on-call load grows with every client you add.
Option 3 — A managed instance per client
This keeps the per-client isolation of Option 2 and removes the operational tax. Each client still gets their own instance, database, and credentials. The patching, daily off-site backups, monitoring, and 24/7 chat are run for you, at €9 per instance. You keep the client relationship and the margin, and you stop being the person who gets paged when a midnight update breaks a workflow. For most small agencies, this is the one that scales.
Three apps that resell well
n8n rarely travels alone. Clients who automate also need somewhere to store the data those workflows touch, and often a way to run their own campaigns. Three open-source tools resell cleanly on the same flat €9 model: n8n for automation, Baserow for client databases, and Mautic for marketing.
Bundle the three and the wholesale cost is €27 per client a month. Retail that stack at €120–180 and the per-client margin holds well above the €9 floor on each app. The point is not the exact mix; it is that every app on the catalogue carries the same predictable cost.
n8n — the automation core
n8n is the engine: it connects a client's apps and runs the workflows that move their data. Managed at €9 per client, isolated per instance, it replaces a per-task automation SaaS whose Team plan starts at $69 a month. It is the anchor of the bundle.
Baserow — the client database
Workflows need somewhere to read from and write to, and a shared spreadsheet does not hold up. Baserow gives each client their own database instance, also at €9, so the data your automations touch lives in something structured and isolated. NocoDB fills the same slot if a client prefers it.
Mautic — the client's marketing automation
For clients who run their own campaigns, Mautic is a self-hosted marketing automation tool that resells on the same flat €9. It gives an agency a recurring line against the per-seat hosted marketing suites a client would otherwise rent.
When this works (and when it doesn't)
The reseller model fits some agencies and fights others. It works when you already own client relationships, want recurring revenue, and would rather not run servers. It does not work when clients demand to build inside the tool themselves, or when nobody on your side wants to own the account.
The model is not for every agency. Two short lists tell you quickly whether it fits.
It fits when:
You already manage ten or more client accounts and want recurring revenue instead of one-off project fees.
Your clients want results and reports, not a login to a tool they will not maintain.
You would rather pay €9 a month than spend your evenings patching servers.
It doesn't fit when:
Your clients insist on building and editing their own workflows inside the n8n editor at scale — that is Embed or OEM territory, with its own license and cost.
A client's volume runs past roughly 10,000 executions a day and needs constant tuning.
You have no appetite to own the client relationship or answer the occasional support question.
How this compares to staying on Zapier
Most agencies arrive here after a Zapier task-overage email. Zapier charges per task, counting every action a workflow takes, and its Team plan starts at $69 a month. n8n charges per execution — one run, one count, however many steps. For multi-step client work, the gap compounds fast.
The trigger for most of these conversations is a Zapier task-overage email. Zapier counts a task every time a workflow completes an action, so a workflow with ten steps spends ten tasks each time it runs.
n8n counts differently. One workflow run is one execution, however many steps it contains and however much data it moves. Run that same ten-step workflow a thousand times a month and you have used a thousand n8n executions against ten thousand Zapier tasks.
Zapier's Team plan starts at $69 a month and meters from there. Self-hosted n8n is free under its license, and a managed instance is €9. Zapier also runs a Solution Partner program, but it pays referral commissions — it is not a way to resell hosting at a margin.
Zapier earns its price for teams that never want to touch infrastructure and keep their volumes low. The moment workflows get multi-step and clients multiply, the per-task meter is what sends agencies looking for n8n.
FAQ
Is it legal to resell n8n to clients?
Running n8n to build and operate automations for clients is permitted consulting under the Sustainable Use License. What the license restricts is selling access to n8n itself, or white-labeling it for money. Give each client their own instance and bill for the managed service, not for n8n access.
Does each client need their own n8n instance?
It is the cleanest model. A separate instance gives each client its own database, credentials, and webhook URLs, so one client's change never touches another's. It also keeps you on the safe side of the license. On DANIAN that is one €9 instance per client.
What does it cost to run n8n for 20 clients?
At DANIAN's flat €9 per app, 20 client instances cost €180 a month wholesale. Retail those at €40–60 per client and you keep €620–1,020 a month in gross margin. The price covers patching, daily off-site backups, monitoring, and 24/7 chat, with no per-execution metering.
What happens if a client runs heavy workflows?
A light instance handles most client work on 2 GB of RAM. Clients pushing past roughly 10,000 executions a day, or running large data through Code nodes, need more headroom. We add capacity at +€9 per vCPU, and only after you approve the change. Nothing scales silently.
Can I move my Zapier automations to n8n?
Most can be rebuilt in n8n, though there is no one-click import. The usual trigger is a Zapier task-overage email. n8n counts one execution per workflow run regardless of steps, while Zapier counts one task per action, so multi-step automations get cheaper to run after the switch.
What if I want to host n8n myself instead?
That path is real. A developer-led team can self-install n8n, or run Coolify on a $24 production-class VPS, and own the patching and on-call. It works if you enjoy that work and have the time. If you would rather not babysit servers at 2am, the managed route is why DANIAN exists.
Does n8n support multi-tenancy out of the box?
No. n8n has no native, first-class multi-tenancy. There is no built-in tenant layer that isolates one client's workflows, credentials, and data from another's inside a single instance. The recommended pattern is hard isolation: a separate instance per client. Logical separation inside one instance is possible but fragile.
Can I white-label n8n for my clients?
Only under n8n's paid Embed or OEM agreement. The Sustainable Use License forbids white-labeling n8n and offering it to customers for money. Without that agreement, n8n branding stays visible in the editor. Most small agencies avoid the issue by operating instances on the client's behalf rather than reselling the interface.
Do I need an n8n Enterprise license to host n8n for clients?
Not for ordinary consulting work. Building and running automations for clients on standard n8n is allowed under the Sustainable Use License. You need an Enterprise or Embed license only for specific cases: white-labeling the editor, embedding n8n inside your own product, or using the enterprise-only source files.
What is the n8n Sustainable Use License in plain terms?
It is a fair-code license n8n has used since March 2022. You may use and modify n8n for your own internal business purposes for free. You may not sell access to n8n itself or white-label it. Consulting, support, and self-hosting are explicitly allowed. It is based on the Elastic License 2.0.
Can my clients log in to their own n8n instance?
They can, if you give them access to an instance you operate for them. The line that matters is whether you are charging for access to n8n itself. Many agencies keep clients out of the n8n editor and deliver results instead, which sidesteps both the license question and the support load.
What is the difference between a shared n8n instance and one instance per client?
A shared instance puts every client's workflows in one process, one database, and one set of system credentials, separated only by tags or folders. One instance per client gives each its own process, database, and credentials. Shared is cheaper to run; per-client is safer, since one misconfiguration cannot cross tenants.
Can I use n8n's Projects feature to separate clients?
Partly. Projects add role-based separation inside one instance, but the tiers cap it: the Pro plan includes 3 projects, Business includes 6, and unlimited projects need Enterprise. Custom roles are Enterprise-only. Past six active clients you outgrow Pro, and you still share one underlying database across every project.
How do I keep each client's data and credentials isolated in n8n?
The reliable way is a separate instance per client. Each instance holds its own credential store, its own database, and its own webhook endpoints, so nothing is shared by default. Inside a single shared instance, separation is logical only, and one wrong setting can expose data across clients.
How does n8n's execution pricing compare to Zapier's task pricing?
They count differently. n8n bills one execution per workflow run, no matter how many steps it has or how much data it moves. Zapier bills one task per successful action, so a ten-step automation burns ten tasks each run. Multi-step and AI-heavy work gets far cheaper on n8n's model.
Is n8n cheaper than Zapier for an agency managing several clients?
Usually, once workflows get multi-step. Zapier's Team plan starts at $69 a month and meters every action as a task. Self-hosted n8n is free under its license, and DANIAN runs a managed instance at €9 per app. The savings widen with each step you add to a workflow.
What is the difference between n8n Cloud and self-hosted n8n?
n8n Cloud is hosted by n8n and billed by execution volume, from €20 a month for 2,500 executions up to €50 for 10,000. Self-hosted n8n is the free Community Edition you run yourself. DANIAN sits between them: self-hosted n8n that we operate for you, at a flat €9 per app.
How much RAM does one n8n instance need?
An idle instance uses around 100 MB. For real client work, 2 GB is the practical minimum and 4 GB is comfortable. RAM, not CPU, is the limit. A single Code node processing a large response can briefly use several hundred megabytes, so size for the heaviest workflow a client runs.
Who fixes a client's n8n workflow if it breaks overnight?
On a self-managed instance, that is you, at 2am. On DANIAN, we patch, monitor, and back up every instance, and a human answers chat around the clock. We usually catch problems before the client notices. The on-call burden is the main reason agencies move off running their own servers.
How should an agency bill clients for n8n hosting?
Bill for the managed outcome, not for n8n access, which also keeps you clear of the license. A common structure is a flat monthly retainer per client that bundles the instance, your setup, and ongoing changes. With a €9 wholesale cost, a €40–60 retail price leaves healthy recurring margin.
When does reselling managed n8n stop making sense?
When clients want to log in and build their own automations at scale, you have crossed into product territory and should look at n8n's Embed or OEM license. It also stops fitting if a client's volume needs constant tuning, or if you have no appetite to own the client relationship.
Conclusion — what to do this week
If a client has already asked you to host their n8n, the move this week is small: pick one client, give them their own managed instance, and price it as a flat monthly retainer. Run it for a month before you productize it across the rest.
Keep each client isolated. Bill for the managed service, not for n8n access. Size the instance for the client's heaviest workflow, and add capacity only when they approve it.
The unglamorous setup — one isolated instance per client, operated by someone else — is the one that scales without turning your agency into a hosting company.
When you are ready to set one up, ask about partner setup and we will walk through the first client with you.
Sources
n8n — Sustainable Use License — https://docs.n8n.io/sustainable-use-license/
n8n — Plans and pricing — https://n8n.io/pricing/
n8n — Hosting documentation — https://docs.n8n.io/hosting/
n8n — Embed / OEM documentation — https://docs.n8n.io/embed/
n8n — GitHub repository — https://github.com/n8n-io/n8n
Zapier — Plans and pricing — https://zapier.com/pricing
