n8n is unusual among self-hosted applications in not publishing a minimum server specification. Its hosting documentation talks about modes, queues and pruning rather than gigabytes, and the honest sizing answer is that n8n's footprint is set by its executions, not by its code. This guide, with n8n's documentation and price list checked on 1 November 2026, sizes a self-hosted n8n for three workloads on MassiveGRID PaaS, built on Virtuozzo Application Platform (formerly Jelastic), and puts the monthly cost beside the n8n Cloud plans.
What n8n's documentation does and does not say
n8n's scaling documentation describes the choices that decide a server's size without giving a size. n8n can run in different modes, and queue mode provides the best scalability, running n8n across multiple worker processes with Redis as the queue and PostgreSQL as the database. At scale, the docs say, the configuration has to change to keep performance good: execution data should be saved selectively and pruned to protect database performance, concurrency can be limited, binary data can be moved to external storage, and memory issues have a troubleshooting page of their own.
What the docs leave to the operator is the arithmetic: a workflow that runs every minute, calls two APIs and writes a row is a few hundred megabytes of Node.js at rest and a small database; a workflow that receives webhook bursts, processes files and feeds an LLM is a different machine. The practical minimums below come from running the package, not from n8n, and are labelled as such.
n8n on MassiveGRID PaaS: one-click install, about $29.52 a month for the Starter configuration (Node.js application server, 1 GiB; PostgreSQL database, 512 MiB) at $0.003372 per cloudlet-hour, before volume discounts, in New York, London, Frankfurt or Singapore. 14-day free trial, no credit card. See the sizing table.
Three workloads and what each needs
Personal automations and testing: a handful of workflows on schedules and a few webhooks. A 1 GiB Node.js node and a 512 MiB PostgreSQL are comfortable; execution data pruning at seven days keeps the database small.
A team running workflows all day: dozens of active workflows, webhooks from a CRM or a form tool, some binary data. 2 GiB for n8n and 1 GiB for PostgreSQL, with execution data saved only for failures and pruned, and Redis added when the first webhook burst causes queued executions.
High-volume webhooks and worker mode: hundreds of executions a minute, AI steps, files. This is queue mode: a main n8n instance for the editor and webhooks, worker instances that take executions from Redis, and a PostgreSQL sized for the execution table. The queue mode post shows the layout and the trigger that adds workers.
The three configurations on PaaS and their cost
The n8n package on MassiveGRID PaaS installs the Node.js application server running n8n, a PostgreSQL node and persistent storage, with authentication and a certificate. Cloudlets are 128 MiB RAM plus 400 MHz CPU at $0.003372 per cloudlet-hour, $2.46 per cloudlet-month; n8n's Sustainable Use License costs nothing for internal business use, and executions are unlimited.
| Configuration | Topology | Cloudlets | Per month (730 h) | Executions |
|---|---|---|---|---|
| Starter | n8n 1 GiB, PostgreSQL 512 MiB | 12 | $29.52 | Unlimited |
| Standard | n8n 2 GiB, PostgreSQL 1 GiB | 24 | $59.04 | Unlimited |
| Production | Load balancer, 2 × n8n 4 GiB, PostgreSQL 2 GiB | 84 | $206.64 | Unlimited |
These are ceilings for an always-on environment before volume discounts; the invoice reflects the cloudlets actually used each hour, and an n8n that is busy for an hour a day is billed accordingly. The PaaS pricing calculator prices other topologies.
Beside the n8n Cloud price list
n8n's pricing page on 1 November 2026 listed, billed annually, a Starter plan at €20 a month with 2,500 workflow executions, a Pro plan at €50 a month with 10,000 executions, an Enterprise plan on request, and a Business plan at €667 a month with 40,000 executions that is self-hosted only, with its cloud version on a waitlist. Active workflows are unlimited on the listed plans; monthly-billed prices were behind a toggle and are not quoted here. The community edition is described simply as a standard self-hosted version available on GitHub.
The comparison is therefore between a quota and a server. 2,500 executions a month is about 83 a day, which a single scheduled workflow running every fifteen minutes exhausts on its own; the Starter configuration above runs that workflow and any number of others for $29.52. At 10,000 executions, the Pro plan's €50 buys roughly what the Standard configuration's $59.04 buys with no ceiling. Past that, the Cloud list jumps to Enterprise pricing on request, while the Production configuration at $206.64 carries hundreds of thousands of executions, as the 200,000-execution cost model works through.
What Cloud includes that the server does not is the operations: upgrades, backups, monitoring and support are n8n's job on Cloud and yours on a server. On the platform the first three are a container redeploy, the Database Backup/Restore add-on and the dashboard's statistics; the fourth is whoever owns the environment.
What to configure on day one
Four settings decide whether the Starter configuration stays enough. Execution data: save successful executions only when you need the audit trail, save failures always, and prune at a retention you can defend; the execution table is what grows. Timezone and webhook URL in the environment variables, so schedules fire when you think and webhooks resolve to your domain. Binary data to the filesystem rather than the database as soon as workflows move files. And credentials encryption key backed up outside the environment, because a restored database without it is unreadable.
When the environment is in regular use, watch two graphs in the dashboard: the n8n node's memory, which tells you when to raise its cloudlet limit, and the database's disk, which tells you whether pruning is working. Both are the signal to move to the next row, and both are the signal the Cloud plans hide behind a quota.
Frequently Asked Questions
What are n8n's minimum server requirements?
n8n does not publish a hardware minimum; its hosting documentation covers modes, queue mode, execution data pruning and memory troubleshooting instead. In practice a 1 GiB Node.js node with a 512 MiB PostgreSQL runs personal and small-team workloads, which is the Starter configuration on MassiveGRID PaaS at about $29.52 a month.
How much does n8n Cloud cost compared with self-hosting?
On 1 November 2026 n8n Cloud listed €20 a month for 2,500 executions (Starter) and €50 for 10,000 (Pro), billed annually, with a self-hosted-only Business plan at €667 for 40,000. A self-hosted n8n on MassiveGRID PaaS costs about $29.52 to $206.64 a month for the three configurations and has no execution limit.
When does n8n need queue mode?
When executions queue behind each other: webhook bursts, long AI or file steps, or hundreds of executions a minute. n8n's docs call queue mode the most scalable setup; on the platform it is the Production configuration, a main instance with workers that scale on a trigger and Redis between them.
n8n without an execution meter
The n8n package on MassiveGRID PaaS installs n8n with PostgreSQL and persistent storage in one click, from about $29.52 a month for the Starter configuration, with unlimited executions and room for queue mode. 14-day free trial, no credit card.
n8n on PaaS