Choosing n8n hosting means choosing who operates the application, its database, its queue and its recovery process. Execution counts matter, but they do not tell you whether a host can run your largest workflow or restore the credentials it needs.
This guide compares the official cloud service, managed self-hosted n8n, a general application platform and a VPS. Prices and product pages were checked on 2 November 2026. DanubeData publishes this comparison and is one of the providers below; the recommendations follow the responsibilities and constraints of each option.
n8n hosting options at a glance
| Option | Published entry point | What constrains usage | Who operates n8n |
|---|---|---|---|
| n8n Cloud | Starter €20/month billed annually | 2,500 monthly executions; concurrency and edition limits | n8n |
| DanubeData managed n8n | From €12.99/month, billed hourly | Resource profile, storage, worker concurrency and edition | Managed application stack; customer owns workflows |
| Elestio | Advertised from US$11/month | Selected server/cloud capacity and configuration | Managed hosting; confirm exact selected deployment |
| Railway | Usage pricing; Hobby US$5 monthly minimum includes US$5 usage | CPU, memory, volumes, plan and n8n configuration | Customer operates n8n on Railway infrastructure |
| Self-managed VPS | Server quote plus backups and operating time | Server capacity, software and licensing | Customer |
These are entry points, not equivalent capacity quotes. The n8n Cloud figures use an annual billing commitment. Currency conversion, tax, external API/model charges and optional software licenses are separate. DanubeData prices exclude VAT. An advertised small instance does not establish capacity for a particular number of executions.
n8n Cloud: official SaaS with execution-based plans
The current n8n pricing page says plans have unlimited users and workflows. Starter includes 2,500 monthly executions and five concurrent executions; Pro lists €50/month billed annually with 10,000 executions and up to 50 concurrent executions. Enterprise uses a negotiated execution allowance. The old five-active-workflow and fifteen-active-workflow limits are not the current comparison.
Choose the official service when its hosted feature set and support arrangement meet your requirements. Compare project sharing, retention, environment controls and support on the actual edition. A third-party Community Edition instance does not automatically include the paid collaboration and governance features of n8n's own plans.
Do not select a larger plan from monthly execution count alone. Five slow, memory-heavy workflows can compete for resources more than thousands of short API calls. Test the longest execution, payload size and peak concurrency as well as the monthly total.
DanubeData: a managed self-hosted stack
DanubeData managed n8n starts at €12.99/month, billed hourly, and provisions n8n and a worker with PostgreSQL, Redis, persistent storage and a TLS endpoint in Germany. Hosting uses a resource profile rather than an n8n Cloud execution tariff. This does not grant unlimited capacity: CPU, memory, storage, worker concurrency and available settings follow the selected plan, and n8n licensing still applies.
The operational boundary matters more than the word “managed.” Confirm the selected plan's upgrade policy, recovery retention and support scope before relying on it. A dashboard snapshot is a saved recovery point; it is not a promise of continuous recovery, zero data loss or a successful restore of every external integration.
| Responsibility | What to establish before purchase |
|---|---|
| Provisioning | Application, PostgreSQL, Redis, persistent volumes and HTTPS are part of the managed stack. |
| Workflow ownership | You create the n8n owner account, configure credentials, build workflows and manage connected services. |
| Upgrades | Agree version selection, update timing, compatibility checks and the recovery path if a workflow breaks. Do not assume every upstream release is applied immediately. |
| Snapshots and restore | Snapshot controls and automated snapshot options exist. Confirm retention, restore procedure, copies and who performs a recovery drill. |
| Capacity and incidents | Choose a profile using peak memory and concurrency. Agree which infrastructure alerts support handles and which workflow/business failures you must monitor. |
| Custom software | Confirm community nodes, available modules, code-node settings and system binaries. Managed application hosting is not a general SSH server. |
If the workflow needs an arbitrary OS package, browser runtime or custom worker topology, get that requirement checked first. A VPS may be a better fit when you need control over the complete image and operating system.
Elestio: managed hosting with deployment choices
Elestio's n8n page currently advertises a US$11/month entry and lists backups, monitoring, TLS and updates, with several infrastructure-provider choices. The older US$32 headline is not a useful current minimum.
Obtain a quote for the actual provider, region, server size and recovery policy you want. Ask whether the proposed instance includes a separate PostgreSQL database and worker/queue layout or a different topology. Confirm terminal access and customization boundaries rather than assuming every deployment has the same capabilities. Its broader cloud choice can be useful when you need n8n close to an existing dependency.
Railway: infrastructure convenience with application responsibility
Railway has an n8n deployment guide, image deployments and templates. Its current regions include Amsterdam; describing it as US-only is wrong. Region selection for an application does not by itself establish the location of every backup, log or external service.
The Hobby minimum is US$5/month with US$5 of usage included, rather than a US$5 fee added on top of all resource usage. Quote the whole topology: n8n, database, Redis, volumes, backups and transfer. You still need to validate n8n persistence, authentication, public webhook URL, worker configuration and restore procedures. A successful template deployment is the beginning of operating the app.
A VPS: control in exchange for operational work
A VPS lets you select the image, packages, network policy and backup tooling. It can be the right choice for workflows needing custom system binaries or a tightly controlled runtime. It still has memory, disk, CPU and network limits, and n8n's license and edition restrictions still apply.
Include PostgreSQL, optional Redis/workers, TLS renewal, monitoring, security patches and recovery storage in the budget. A low server price does not include the time to diagnose a failed upgrade or restore encrypted credentials. Avoid exposing the editor or database broadly to make initial setup easier. Use the supported authentication controls for your exact n8n version.
How to size a real workload
Compare representative executions rather than a synthetic “50,000 executions” price winner. Record duration, peak memory, largest binary payload, execution-data growth, simultaneous webhook bursts and rate limits of downstream APIs. Include failed executions: saving a large failed payload can be an important part of memory and storage demand.
| Example | What to measure | What the number does not prove |
|---|---|---|
| 1,000 small scheduled executions/month | Peak concurrency, API rate limits and retention | That any smallest plan will handle a large attachment or browser task |
| 50,000 executions/month | Duration distribution, queue wait, worker saturation and database growth | That a fixed resource profile can process them without delay |
| A few large file/AI workflows | Payload sizes, worker/database memory, binary storage and external model cost | That “unmetered executions” includes model tokens or unlimited RAM |
Queue mode and recovery questions to ask
n8n queue mode separates accepting work from executing it: Redis carries execution IDs, workers load data from the database, and the main process coordinates triggers. It is a scaling architecture, not a guarantee that jobs cannot stall. Regular mode also supports concurrent execution; it is inaccurate to say one slow workflow always blocks everything without Redis.
In queue mode, the main process and workers need the correct shared credential encryption key. The official documentation does not support filesystem binary storage with queue mode; verify a supported storage configuration and any edition requirements for your version. A backup plan must preserve the database, necessary encryption material and required persistent data securely. A workflow JSON export alone does not establish recovery of credentials, schedules and execution state.
- Define the tolerated data-loss window and recovery time for the business process.
- Record versions and retain a recovery point before upgrading.
- Restore into an isolated instance with outbound delivery and schedules disabled.
- Check credential decryption and a harmless representative execution before re-enabling effects.
- Use idempotency keys or a durable business identifier so a replay does not send or charge twice.
Use our stuck execution diagnostic guide and n8n migration, backup and recovery guide for the operational checks. For a Romanian buying checklist, see managed n8n versus VPS hosting.
European hosting and connected services
Record the locations and providers for application data, execution logs, backups and support access. Then map the services each workflow calls. A Germany-hosted workflow can send personal data to a CRM, email provider or model API elsewhere. Review the provider's contractual documents and your application's processing requirements; no hosting entry in this table certifies a workflow's GDPR compliance.
Which option fits?
- Official cloud features and support: start with n8n Cloud and compare the execution and edition allowance you actually need.
- A resource-priced managed application in Germany: evaluate DanubeData, including capacity, upgrade and recovery responsibilities.
- A particular infrastructure provider or broader managed-hosting choice: obtain an Elestio quote for the exact topology.
- An existing Railway environment: use its deployment facilities if your team will operate n8n and its data services.
- Custom system tools and full runtime control: use a VPS when your team can own security, updates and restores.
Frequently asked questions
Does unlimited workflows mean unlimited executions?
No. n8n Cloud's current plans distinguish unlimited workflow definitions from a monthly execution allowance. Resource-priced self-hosting can avoid an execution tariff while still enforcing CPU, memory, concurrency and storage limits.
Is n8n free to self-host?
A Community Edition is available, but n8n uses a fair-code licensing model and paid editions add features. Check the current edition and licensing terms for your use, especially if you embed or resell automation services. Hosting and connected APIs also cost money.
Does “managed” mean the host fixes my workflows?
Only if that support is explicitly included. Provisioning the application and database does not establish responsibility for a broken third-party API, expired OAuth credentials or an incorrect business rule.
Compare the current DanubeData managed n8n offering with a representative workflow and a written recovery requirement before choosing a profile.
