Self-hosting n8n can give you more control over where workflows run. It also gives you responsibility for keeping them available. The honest monthly cost includes infrastructure, any applicable software licence, backups, monitoring, maintenance time and the consequences of a failed process.
The software price is one line in the budget
A small server invoice does not tell you whether the deployment is dependable. If a workflow handles leads, customer enquiries or operations, somebody needs to notice a failure and recover the work. A low-cost installation with no recoverable history can become expensive precisely when the business needs it.
Use the worksheet below before choosing between self-hosting and a managed n8n plan. It is a planning tool, not a provider quote or a claim that every deployment needs the same infrastructure.
Separate your cash cost from your time cost
Create two subtotals. The cash subtotal includes the server, storage, database, backups, monitoring and any paid features required by your use case. The time subtotal includes installation, routine updates, incident response and restore rehearsals. Do not hide maintenance because the developer is also the business owner.
Check the current n8n pricing page for managed-plan allowances and features. Paid plans are based on workflow executions rather than simply counting each step. Check the applicable licence for your self-hosted use case too. A Community deployment and a deployment requiring paid capabilities are not interchangeable budget assumptions.
For external dependencies such as model APIs, email delivery or a CRM, record those costs separately. Moving the workflow engine to your own server does not make those services free. Their rate limits and outages can still interrupt the process.
- Infrastructure: compute, persistent storage and database, with expected traffic and memory needs.
- Recovery: backup storage, retention and time to rehearse a restore.
- Operations: monitoring, alerting, upgrades and credential management.
- Dependencies: model usage, email, paid APIs and connected business tools.
- People: named ownership for routine changes and an incident outside normal working hours.
A worked budget—with invented inputs clearly labelled
Here is an illustrative worksheet, not current market pricing: compute €18, storage and backups €7, monitoring €5, and two maintenance hours valued at €40 each. Those assumptions produce €30 in recurring infrastructure costs and €80 in time cost, for a €110 monthly operating baseline. The arithmetic is 18 + 7 + 5 + (2 × 40).
That baseline excludes model API usage, paid n8n features, migration work and incidents. If your team already operates a suitable platform, some incremental costs may be lower. If nobody has hosting experience, setup and recovery costs may be higher. Replace every input with your own quote and realistic ownership estimate.
Compare that total with the managed plan that actually meets your execution allowance and feature requirements. Do not compare a production self-hosted system with the cheapest managed tier if the tier does not support your workload. Equally, do not assume the largest managed tier is required just because the workflow has many steps.
Prove you can recover before calling it production
A backup is only part of recovery. You need the workflow data, the configuration and a safe way to restore any secrets or credential-related material your deployment requires. Follow the current documentation for your installation rather than improvising from an old container tutorial.
Restore into an isolated test environment. Keep outbound actions disabled so restored workflows do not send old messages or repeat business transactions. Confirm that an operator can inspect the restored workflows, reconnect approved dependencies and identify what was incomplete at the point of failure.
Then measure recovery time. If the business cannot tolerate a day without processing enquiries, a recovery process that takes a day is not adequate even if the backups exist. Decide whether better operations, a managed service or a simpler workflow is the appropriate answer.
Retries and history affect the real workload
Count the work that happens outside the happy path. A workflow may retry when a service is unavailable, wait for review or be triggered again by a duplicate event. Record how your installation and plan count those executions and how much history you intend to retain.
Use a stable submission identifier for actions that must not repeat. For example, retrying a lead-processing job should not create another CRM record or send the same confirmation twice. Making retries safe is application design work; buying a larger server does not solve it.
Track the successful business outcome as well as the run count. If 500 executions lead to 300 completed requests and 200 unresolved failures, the low infrastructure cost is not the useful result. Include manual recovery time when you calculate cost per completed request.
When a managed plan may be the better purchase
A managed plan deserves serious consideration when your team lacks a reliable operating owner, the workflow is business-critical or maintenance would distract from more valuable work. Self-hosting deserves consideration when control over deployment is useful and you can support it properly.
Is self-hosted n8n free? Some software use may be available without a subscription under the applicable terms, but a dependable deployment still has operating costs. Hosting or embedding for clients also needs a licence check against the actual use case. Read the official guidance rather than treating a headline as permission.
Set a review point after the first month of real use. Compare expected executions, infrastructure cost, maintenance time and unresolved failures with what actually happened. Keep the hosting choice that supports the business reliably, not the one that produced the smallest invoice on launch day.