Best n8n hosting options (2026)
What actually decides this is who owns Postgres, queue mode, and the 2am upgrade when a workflow stalls, not which host is cheapest.
The short answer
A self-managed VPS with Docker Compose, Postgres and queue mode is the best n8n hosting for production. n8n Cloud is right until a workflow touches a private database or runs often.
n8n's own docs are blunt about the default: SQLite ships out of the box and isn't recommended once you run queue mode. Queue mode needs PostgreSQL, with n8n supporting versions 17 and 18 as of July 2026, plus Redis as the message broker. Hosting is really a decision about who carries that stack.
We run n8n for Little Tree Confections' meeting-to-action pipeline. It reads the client's live task structure, and that's the kind of job that decides this ranking. It touches real data, so it runs on infrastructure we control.
Key takeaways
3 things that decide this
- 01A self-managed VPS with Docker Compose is the best n8n hosting option once a workflow touches a private database. Railway gives you the same stack with no compose file to write.
- 02We ranked five options on data residency, what breaks the billing model, who owns the upgrade, and the path to queue mode, checked on 11 November 2026.
- 03Before you self-host, name the person who owns upgrades and backups. Without one, n8n Cloud or a managed host is the safer pick.
How we ranked these n8n hosting options
Verified
We ranked five ways to run n8n on what a production workflow needs, not on sticker price. A workflow that touches customer data, runs on a schedule, or has to survive a failed step needs Postgres. Past a handful of concurrent users, it needs queue mode with Redis workers too.
For every entry we read n8n's own documentation and the host's own pages on 11 November 2026, linked in each entry. We host n8n for a client in production, and that experience sets the criteria.
Hashlogics publishes this page. We build custom software and are not one of the tools ranked here. No company paid for its place.
- Data residency
- Whether the workflow's data ever leaves infrastructure you control.
- What breaks the billing model
- Execution count, active workflows, or nothing at all.
- Who owns the upgrade
- You, a platform's build pipeline, or nobody until something breaks.
- Path to queue mode
- How far you get before Redis and worker processes stop being optional.
The five at a glance
Read from n8n's documentation and each host's own pages on 11 November 2026.
| Option | Who owns the database | Billed on | Best moment to pick it |
|---|---|---|---|
| n8n Cloud | n8n | Executions per month | You want zero infrastructure and no private data |
| Railway (managed platform) | You, provisioned automatically | Compute and database usage | You want Postgres and Redis without writing the compose file |
| Self-managed VPS + Docker | You, fully | The server, flat | The workflow touches a private database |
| Elestio (managed) | Elestio, on a dedicated VM | Hourly, per deployed service | You want self-hosted n8n without patching it yourself |
| Kubernetes + Helm chart | Your platform team | Cluster capacity, flat | You already run workloads on Kubernetes |
The ranking
Order reflects how many teams each one is right for, not raw capability. Each entry was checked on n8n's docs and the host's own pages on 11 November 2026.
Postgres, queue mode, and the data stay where you put them
This is where every production instance we run ends up. n8n's own guidance recommends Docker for self-hosting because it isolates the app from operating-system quirks and makes database management simpler. Docker Compose is the documented path for a setup with a real database and extra services, rather than one process running alone.
Best for
- Workflows that read or write a private database
- Teams that need queue mode and are willing to run Redis
- Anyone who needs a clear answer to "where does the data live"
Not for
- Teams with nobody willing to own an upgrade schedule
- Low-volume workflows against public SaaS tools only
- Database
- Postgres 17 or 18, self-managed
- Scaling path
- Queue mode: Redis broker, separate workers
The Docker Compose stack, provisioned for you
The n8n with workers template on Railway's marketplace deploys Postgres, Redis, and worker processes together with no compose file to write. It is the same architecture as the option above, minus the afternoon spent debugging environment variables.
You still own the data and the upgrade decision, since the containers run in your Railway project. What you give up is the fine-grained server control a bare VPS gives you. In exchange, the bill scales with usage instead of staying flat.
Best for
- Teams who want Postgres and queue mode without writing infrastructure
- A first production instance before hiring anyone to own a VPS
Not for
- Workloads where a predictable flat cost matters more than convenience
- Teams that already have a VPS or Kubernetes habit
- Setup
- One-click template, Postgres and Redis included
- Billed on
- Compute and database usage
No infrastructure, execution-based billing
Pick this to get moving before any of the hosting questions matter. There is nothing to patch and nothing to back up. n8n's pricing counts one execution per whole workflow run rather than per step, so a small number of long workflows costs little.
The trade is that a private database now sits behind a third party's infrastructure. A workflow that runs often will also hit the execution ceiling sooner than it feels expensive. Neither is a flaw. Both are reasons to move once the workflow gets serious.
Best for
- A first workflow, or a prototype before anything is decided
- Automation that only touches public SaaS tools
Not for
- Workflows that read or write a private or customer database
- High-frequency workflows, where execution billing adds up fastest
- Runs as
- Fully managed, n8n's infrastructure
- Billed on
- Executions per month
Self-hosted n8n on a dedicated VM, with the patching done for you
Elestio runs open-source n8n on a dedicated virtual machine in the cloud you choose, including Hetzner, DigitalOcean, AWS, Azure and OVHcloud. Backups, SSL, software and OS updates are handled for you, and billing is hourly per deployed service.
You get most of the self-hosted benefit without owning the upgrade. The trade is that a third party still runs the box, and queue mode with separate workers is yours to configure. Elestio is based in Dublin, Ireland.
Best for
- Teams that want n8n on its own server without patching it
- A first self-hosted instance with backups already set up
Not for
- Teams whose policy rules out a third party with server access
- Setups that need fine-grained control over queue-mode workers
- Runs on
- Dedicated VM, your choice of cloud
- Upgrades
- Handled by Elestio
For teams already running everything else this way
n8n publishes an official Helm chart. It is the right choice for one situation. A platform team that already runs its other services on Kubernetes, and does not want n8n to be the exception.
It ranks last only because it is the most machinery for a tool that does not usually need it. A single client-facing automation almost never justifies a cluster. Adopt this path because you already have one, not because n8n asked for it.
Best for
- Platform teams standardising every workload on Kubernetes
- Multiple n8n instances across environments, managed the same way
Not for
- A single production workflow with no cluster already in place
- Teams without someone who already operates Kubernetes
- Deployment
- Official Helm chart
- Fit
- Existing Kubernetes platforms only
- n8nThe editor and the API.
- PostgresRequired past SQLite. Workflows and credentials live here.
- RedisThe queue-mode broker, once volume needs it.
- WorkersSeparate processes that actually run each workflow.
- UpgradesThe recurring cost every hosting choice is really about.
The host you pick decides who is responsible for each of these, not whether they exist.
When none of these is the answer
A workflow that needs guaranteed execution, retries with backoff, and a durable record of every attempt has outgrown n8n itself. Hosting will not fix that. n8n has no durable replay: a failed run in the middle of a long workflow gets re-triggered, not resumed. That is fine for most automation and wrong for anything where a missed step has real consequences.
If that is your workflow, the fix is not a bigger server. It is a job queue built for durable execution, with n8n or a code service sitting in front of it.
That's the kind of automation we build and run for clients, with the retries and recovery designed in. See business process automation.
- 01If the answer to a failed run has to be provably recoverable, this is the wrong tool regardless of host.
- 02If the workflow runs a handful of times a day against public SaaS only, n8n Cloud is simpler than any self-hosted option.
- 03If nobody on the team will own upgrades, budget for someone who will before you self-host anything.
n8n hosted and running unattended, in production
Not sure which host fits your workflow?
Tell us what the workflow touches and how often it runs. We will name the setup we would use and why. Scoping calls cost nothing. Every build then starts with a fixed-fee Blueprint, credited in full against the build, and if the plan isn't one you'd act on, you don't pay for it.
Questions buyers ask
01What is the best way to host n8n?+
A self-managed VPS with Docker Compose, Postgres and queue mode is the best way to host n8n for production. n8n's own docs recommend Docker for most self-hosting. n8n Cloud is the better start for workflows that touch only public SaaS tools.
02Is n8n Cloud good enough for production?+
Yes, for the right workload. It suits automation that never touches a private database and runs infrequently enough that execution-based billing stays cheap. Once either condition stops being true, a self-managed instance is the better production fit.
03Can I start on n8n Cloud and move later?+
Yes. Workflows export as JSON and the logic inside code nodes is ordinary JavaScript, so both port to a self-hosted instance. What does not port automatically is the credentials and connections, which have to be re-added on the new host.
04How do I compare n8n hosting options fairly?+
Run the same workflow on each option with Postgres behind it and two parallel executions. Then check who patches n8n, where the data sits, and what the bill does when executions double.
05Do I need Redis and queue mode from day one?+
No. n8n runs fine in regular mode for low-to-moderate volume. Queue mode becomes necessary once concurrent executions start backing up behind each other. n8n's own docs put a single instance at up to 220 workflow executions per second, with more instances in queue mode scaling past that.
06Why does self-hosting outrank the managed platforms here?+
Because most of the workflows worth automating touch a database somebody cares about. Once that is true, data residency stops being optional. The ranking follows what the workflow needs, not what is fastest to click through.
07What does self-hosting cost in effort, not money?+
A container, a Postgres database, and someone who owns upgrades. Scaling past one instance adds Redis and worker processes. The honest cost is ongoing attention, not a one-time setup.
08Why isn't a generic VPS provider like DigitalOcean or Hetzner listed separately?+
Because they are the same option as the self-managed VPS entry. n8n's Docker Compose setup runs the same way on any provider that gives you a server and a shell. The choice that matters is Docker Compose versus a managed platform versus Kubernetes, not which company sells the underlying box.
09How were these n8n hosting options ranked?+
We ranked five options on data residency, billing model, upgrade ownership and the path to queue mode, checked on 11 November 2026. Hashlogics publishes this page and builds custom software, not one of these hosts. No company paid for its place.
Related reading
- n8n, reviewed from production use →What it does well and where it stops being the right tool.
- What to use when you outgrow n8n →Temporal, Windmill, and Make, and when the honest answer is to stay.
- Hire n8n developers →Engineers who treat workflows as code, not a canvas nobody reviews.
- Business process automation →How we build automation that runs unattended.
