When UK businesses should self-host n8n for AI workflows

Self-host n8n when a demonstrated infrastructure requirement justifies the operational work. For UK small and mid-sized businesses, a free licence alone is a weak reason to move.

7 min read
Read with AI

Open in

ChatGPT Claude Perplexity

This page

Copied to clipboard
A small UK software company’s modest operations room on a rainy weekday morning: a systems engineer in profile, face turned away, checks a compact rack-mounted workstation and a printed infrastructure

Our view is that UK small and mid-sized businesses should choose n8n Cloud unless a specific infrastructure requirement justifies self-hosting and someone can own the resulting service. n8n distinguishes managed Cloud deployment from self-hosting with control over configuration, upgrades and scaling. That control becomes valuable when it solves a demonstrated constraint. A smaller software bill alone is insufficient reason to take responsibility for production infrastructure.

The hosting decision starts with an operational requirement

For a UK business, the useful question is which requirement the proposed deployment must satisfy. That might be an agreed infrastructure location, a particular network arrangement, control over upgrades or a workload that needs different resources.

Self-hosting is a credible answer because n8n explicitly supports running on your own infrastructure, including on-premises and private-cloud deployments. It is not evidence that every desired integration, security feature or commercial use is included in the free edition. Establish those requirements before choosing the licence.

Consider a hypothetical British distributor using AI to prepare product descriptions for staff approval. If a Cloud pilot meets its processing deadlines and access requirements, self-hosting needs a stronger justification than an unused server. The business should first improve the workflow and measure the result.

A mid-sized engineering firm might reach a different conclusion if it needs an approved internal deployment and already operates production applications. Its existing team makes self-hosting more plausible, but the decision should still name the person responsible for n8n, their cover and their maintenance budget.

Neither example establishes where n8n Cloud stores UK customers’ data. The supplied evidence does not resolve Cloud hosting location, contractual data-handling terms or the full edition entitlement matrix. Obtain those details before making residency or feature availability decisive selection criteria.

When self-hosting n8n makes sense
Start with n8n Cloud, then self-host only if a tested requirement remains unmet and someone is funded to operate the service.

Control is useful only when someone operates it

n8n describes self-hosting as providing finer control over upgrades, configuration, security and custom scaling. Our position is that each requested control should have an owner and a measurable purpose.

If the reason is upgrade timing, name who tests releases and authorises deployment. If the reason is recovery, demonstrate a restore. If the reason is performance, identify the unacceptable delay and reproduce it with representative inputs.

For AI workflows, separate the automation platform from the model service. In an illustrative document-processing design, n8n receives a document, sends selected content to a model endpoint, records the result and routes it for approval. Putting the n8n instance on a UK server does not determine where that model endpoint processes the content.

The following is a proposed responsibility allocation to settle before production, rather than a statement of contractual service coverage.

| Responsibility | Decision to settle |

| --- | --- |

| n8n infrastructure | For Cloud, confirm the supplier’s service scope. For self-hosting, assign patching, monitoring, capacity and recovery ownership. |

| Model endpoint | Record its operator, permitted inputs, processing location and failure behaviour. |

| Workflow credentials | Assign ownership of permissions, rotation and emergency revocation. |

| Execution records | Decide what to retain, who may inspect it and when it should be deleted. |

| Business output | Name the person who approves consequential actions and handles incorrect results. |

Self-hosting should not proceed while these responsibilities depend on the continued availability of one enthusiastic employee.

Compare the full operating cost

The strongest financial argument for Cloud is the work it removes. n8n describes it as offering less infrastructure and maintenance work. The strongest financial argument for self-hosting is conditional on whether your organisation can provide the required service economically.

The pricing evidence supplied on 28 September 2026 contains euro figures, but its extracted billing selector does not establish the commitment attached to each displayed amount. It also does not establish VAT treatment. Those figures cannot support a reliable GBP monthly comparison.

Instead, compare both options over the same operating period and include the following costs.

| Cost component | n8n Cloud assessment | Self-hosted assessment |

| --- | --- | --- |

| Software | Confirm the plan, execution allowance, support and billing commitment. | Confirm whether Community meets requirements or a paid edition is needed. |

| Infrastructure | Establish included resources and any upgrade required. | Price compute, database, storage, backups and network traffic. |

| AI processing | Price the selected model service and expected usage. | Price remote model usage or the infrastructure and operation of a local model. |

| Implementation | Include workflow design, credentials, testing and staff training. | Include the same work plus platform deployment and operational documentation. |

| Ongoing operation | Budget for workflow maintenance and supplier escalation. | Add infrastructure maintenance, monitoring, recovery exercises and incident cover. |

| Exit | Establish export, reconfiguration and transition work. | Include data transfer, service handover and secure decommissioning. |

These are quotation requirements, not measured costs. A free Community licence is a documented option; free administration is an assumption.

For a UK budget, ask suppliers to state currency, VAT treatment, billing commitment and support hours. If a partner will operate the installation, require a written division between hosting support, n8n maintenance and workflow repair.

Test the bottleneck before changing the hosting

The supplied n8n pricing page lists Starter at 5 concurrent executions and Pro at 20 concurrent executions. n8n’s Cloud concurrency documentation explains that regular-mode limits apply to production executions started by triggers or webhooks, with excess work queued.

Those numbers describe simultaneous work. They do not show how quickly a particular AI workflow completes.

Before migrating, measure queue delay, execution duration, peak memory use and failures with realistic inputs. Include slow model responses and the largest documents the business expects to process.

There is a credible retain-and-improve option. n8n’s Cloud data-management guidance recommends processing smaller batches and using sub-workflows that return limited results. Test whether those changes meet the deadline before buying more infrastructure.

If scaling remains necessary, self-hosted queue mode separates incoming triggers from execution work. The main instance passes an execution identifier through Redis; a worker retrieves the workflow from the database and writes back its results. Workers must share the main instance’s encryption key to access stored credentials, according to the queue-mode documentation.

That architecture also has constraints. The same documentation says queue mode does not support filesystem storage for binary data. A document-heavy AI workflow therefore needs its storage design and edition requirements checked before migration.

Self-hosting is not the only route to queue mode. n8n documents queue-mode availability for Cloud Enterprise, enabled by contacting the supplier. Compare that option if the workload needs scaling but the business wants managed infrastructure.

Editorial analysis

Our recommendation is to make Cloud the starting preference when the business lacks an established application-operations team. Its managed deployment model leaves fewer infrastructure duties competing with workflow development.

Self-hosting earns its place when an important requirement remains unmet after a representative Cloud trial, the proposed installation demonstrably meets it, and the business funds continuing operation. A partner-operated deployment can satisfy the ownership condition, provided the contract covers recovery, upgrades, escalation and exit.

The counterargument is that an experienced internal platform team may find another managed service unnecessary. That is reasonable where n8n fits its existing operating practices. The deciding evidence should be a working pilot, a recovery test and a costed owner, rather than confidence that installation will be easy.

Before approving a move, ask for a short decision record naming the constraint, the alternatives tested and the acceptance result. If nobody can state what self-hosting will improve, retain Cloud and spend the effort on workflow quality.

FAQ

Is self-hosted n8n free for a UK business?

n8n says an installation without a licence key runs as the free Community edition. That does not establish that Community includes every feature your business needs. Budget separately for infrastructure, administration, support and model processing, and confirm the applicable licence terms.

Will self-hosting keep all AI workflow data in the UK?

Self-hosting lets you choose where the n8n instance runs, including your own infrastructure. It does not, by itself, establish the location of every connected service. In a workflow that sends documents to an external model endpoint, check that endpoint and the storage, logging and backup destinations separately.

Should we self-host when n8n Cloud starts queueing work?

First establish whether the delay breaches a business requirement and whether workflow changes resolve it. Cloud deliberately queues production executions above the applicable concurrency limit. Compare optimisation, a different Cloud plan and self-hosting using the same workload and deadline.

What should we prove before moving a production AI workflow?

Our minimum recommendation is a representative load test, a successful restore and a named operational owner with cover. Test credential recovery, model failures, human approval and the handling of interrupted work. For a queue-mode design, also validate the shared encryption key and binary-storage requirements.

Sources

Published n8n Cloud concurrency limits. Source: n8n Plans and Pricing
Starter and Pro simultaneous production execution limits from the supplied pricing evidence, which are not throughput benchmarks or monthly execution allowances.