Self-Hosted n8n Cost: What the Free Software Still Requires

The cost of self-hosted n8n is infrastructure plus operating labor plus every service called by a workflow. Budget the server, database, storage, backups, monitoring, upgrades, security work, incidents, and staff support. Add model APIs and other paid integrations separately. Then divide the monthly total by accepted business outputs.

n8n offers a standard self-hosted Community Edition, while paid self-hosted plans cover different features and support. That product structure was checked on August 12, 2026 on the official n8n pricing page. “Self-hosted” does not always mean “no n8n license cost,” so confirm the edition and required features before estimating.

The cheapest server is rarely the largest cost. For a small deployment, a few hours of maintenance or one recovery incident can exceed the infrastructure invoice.

Which infrastructure costs belong in the budget?

List each component required to restore the service, not only the machine running the application.

At minimum, consider:

• Application compute and any worker processes.
• A production database and its storage.
• Backup storage in a separate failure domain.
• Network traffic, domain, and certificate management.
• Monitoring, alert delivery, and log storage.
• A staging environment for upgrades.
• Secrets storage and identity controls.

Your design may combine some components. That can reduce invoices, but it does not remove the function. If the database and application share one machine, the backup and recovery plan matter more.

Use current quotes from the providers you will actually buy. Record region, billing term, included traffic, taxes, and quote date. Hosting prices change, so a universal dollar figure would be misleading.

How much operating labor should you include?

Price the time required to keep workflows running. Common duties include installing updates, testing changed nodes, reviewing failed executions, managing disk use, rotating credentials, helping users, and restoring service.

Use a loaded hourly cost for employees or the quoted rate for outside support. Do not price the work at zero because a founder handles it after hours.

Track labor for a month during the pilot. Separate planned maintenance from incidents and user requests. This gives you a usable baseline and exposes whether one workflow creates most of the work.

The AI agent maintenance cost guide provides a broader maintenance worksheet for model and agent changes.

What should the monthly worksheet contain?

The amounts below are assumptions for an example deployment. They are not hosting prices or a claim about what n8n requires.

If the deployment produces 2,000 accepted records, this example costs $0.34 per accepted record. If only 1,000 records pass validation, the cost doubles. This is why the denominator must be accepted work rather than executions.

Replace every line with an invoice or time record after launch. Keep model usage variable so a volume change does not hide inside “hosting.”

How do workflow design choices affect cost?

Polling can run a workflow even when there is no new work. Large binary files can increase storage and backup time. Long retention can fill disks. Loops and retries can add compute load and external API charges.

Measure:

• Executions by workflow and trigger.
• Successful, failed, and retried runs.
• Peak concurrency and queue delay.
• Database and binary-data growth.
• Log volume and retention.
• External calls per accepted output.

Do not delete all execution history just to reduce storage. Keep enough evidence to investigate failures and meet your operational needs. Set a documented retention period and test the cleanup job.

The n8n, Zapier, and Make comparison explains how execution-based economics differ from task and credit billing.

What do backups and recovery really cost?

A backup is useful only if the team can restore it. Budget storage, encryption, access control, monitoring, and restore tests.

Define a recovery point and recovery time based on business impact. These are business choices, not universal numbers. A workflow that sends optional internal summaries can tolerate more downtime than one that moves paid orders.

Test a restore in an isolated environment. Verify workflows, credentials, database records, and required files. Record how long it took and which steps failed. That test time belongs in maintenance.

Also plan for bad workflow changes. Version the workflow export and configuration. Keep a rollback path that does not depend on the broken instance.

How should upgrades be handled?

Do not update a production instance without a tested backup and rollback path. Read release notes, update staging, run representative workflows, and compare outputs.

An upgrade test should include credentials, webhooks, community nodes, custom code, database changes, and external integrations. AI workflows also need output evaluations because a technically successful run can still return unacceptable content.

Choose an update cadence based on security exposure and change tolerance. There is no safe universal interval. Emergency fixes may need a different process from routine feature updates.

Record the version, test evidence, approver, deployment time, and rollback result. This small change log reduces recovery time later.

Which security costs are easy to miss?

Self-hosting transfers control and work to the business. Restrict administrative access, protect the editor, encrypt transport, store secrets outside workflow text, and review who can run or change workflows.

Budget dependency updates, access reviews, credential rotation, audit logs, and incident response. If workflows can send email, change records, or move money, use narrow service-account permissions and approval gates.

Local hosting does not create a private pipeline by itself. Workflow inputs may come from cloud apps. Logs, backups, model calls, monitoring, and notifications may send data elsewhere. The small-business AI privacy checklist helps map the complete data path.

This is operational guidance, not a compliance determination. A business with contractual or regulatory duties should have qualified people review the actual system and data.

When is managed n8n the better buy?

Managed hosting may be cheaper when the team lacks an operator, needs vendor support, or values faster setup over infrastructure control. Self-hosting may fit when the team already operates comparable services or needs deployment control.

Compare both options with the same workload and service expectation. Include:

• Platform or license charges.
• Internal and external operating labor.
• Required support level.
• Backup and recovery responsibility.
• Data location and access needs.
• Expected downtime cost.

Do not compare a managed production plan with an unmonitored server. Those are different services.

How can you keep the cost predictable?

Assign an owner. Set infrastructure and API alerts. Cap retries. Limit concurrency where overload would create duplicate work. Use separate credentials for development and production.

Review monthly cost by workflow. A total server invoice cannot show which automation is wasteful. Pair cost data with accepted output counts and failure records.

Use the LLM API budget guide for model limits and the spending-limits guide for guardrails. Test alerts before relying on them.

Recalculate after major volume, integration, or retention changes. Self-hosting is economical only when the saved subscription cost and added control exceed the full cost of ownership.