Self-Hosted n8n vs n8n Cloud — Performance, Cost, Security (2026)

Source: ai-research/n8n-self-hosting-cost-2026-expresstech.md, ai-research/n8n-blog-no-active-workflow-limits-2026.md, ai-research/n8n-docs-queue-mode-scaling-2026.md, ai-research/devto-n8n-selfhosted-vs-cloud-security-2026.md

The performance, cost, and security trade-offs between self-hosting n8n and n8n Cloud, at scale — not just the entry-level pricing comparison. Resolves the research-agenda question “Self-hosted n8n vs n8n Cloud — performance, cost, and security trade-offs at scale.” Companion to n8n vs Claude Code, which covers the higher-level “n8n or Claude Code” decision and touches self-hosting cost only in passing.

Key Takeaways

• n8n Cloud pricing is execution-capped, self-hosting is not. Cloud tiers: Starter $24/mo(2,500executions,5activeworkflows),Pro$60/mo (10,000 executions, 50 active workflows), Enterprise custom/unlimited. A single workflow triggering every 5 minutes burns 8,640 executions/month — the Starter plan runs dry in 9 days from one workflow. n8n’s 2026 pricing overhaul removed the active-workflow limit from every plan (unlimited workflows/steps/users on every tier now), but the execution cap on Cloud Starter/Pro remains the real constraint at volume.
• Self-hosting costs one to two orders of magnitude less at real usage volumes. Managed one-click platforms (InstaPods, PikaPods): $3-7/mo,unlimitedexecutions,zeromaintenance.RawVPS+Docker(e.g.HetznerCX22):$4-5/mo, full control, you own OS/Docker/n8n updates and SSL renewal. Coolify-on-VPS: $5-8/mo.Fullymanaged(Elestio):$17/mo, close to Cloud Starter pricing but without the execution cap. A concrete worked example: ~7,020 executions/month across four workflows hits Cloud’s Starter cap in 11 days (forcing an upgrade to the $60/moProtier),whilethesameworkloadrunsunmeteredona$3-7/mo self-hosted instance — a $636-684/year difference for one moderate user, multiplying for teams.
• Performance at scale is a queue-mode question, not a self-hosted-vs-cloud question. A single-instance n8n setup (the default) chokes under heavy concurrency — 50+ simultaneous workflow triggers without queue mode risks instability. Enabling queue mode (Redis-backed job queue + separate worker processes) is what actually scales throughput: one documented benchmark run went from ~22 requests/second to ~60.2 req/s (nearly 3x) and cut response time from ~900ms to ~330ms (also ~3x) when scaling from a small number of virtual users to 20, using n8n’s own official benchmarking tool — the same tool n8n uses to benchmark Cloud’s own performance. Self-hosted deployments can enable queue mode; Cloud’s underlying architecture already runs it, but you don’t control or tune it directly.
• Queue mode has real infrastructure requirements. Postgres 13+ is required (SQLite is explicitly not recommended once queue mode is enabled); Redis is required for the job queue; worker concurrency defaults to 10 but n8n recommends 5+ — setting low concurrency across many workers can exhaust the database connection pool and cause processing delays. Multi-main setup (multiple main processes with automatic leader election for high availability) is gated to Self-Hosted Enterprise and Cloud Enterprise plans only — it is not available on Community Edition or lower Cloud tiers.
• Security and compliance cut in favor of self-hosted for regulated/sensitive data, but self-hosted security is DIY unless you pay for it. The compliance argument (GDPR, HIPAA, SOC 2 workloads, customer/patient data) favors self-hosting because you control exactly where data lives and who can access the infrastructure — relevant for any WEO-Marketly-adjacent dental-data use case, though this article does not resolve WEO’s own adoption status (see research agenda for that separate, still-open question). But base self-hosted (Community Edition) puts patching, SSL renewal, firewall rules, and backup integrity entirely on you — “when your SSL cert expires at 2am, that’s on you.” SSO, LDAP, and formal access-control options only ship on the paid self-hosted Business plan or above, not the free Community Edition. n8n Cloud Enterprise, by contrast, includes SSO, LDAP, and audit logs out of the box as part of what you’re paying for.
• The honest trade-off is control vs. convenience, not secure vs. insecure. Neither deployment mode is unconditionally “more secure” — self-hosted gives you data-residency control but shifts patching/operational security work onto your team; Cloud centralizes that operational burden onto n8n but means your data lives on their infrastructure under their access model.

Cost Comparison Table

Option	Monthly cost	Execution limit	Maintenance burden
n8n Cloud Starter	$24/mo	2,500 executions, 5 active workflows	None
n8n Cloud Pro	$60/mo	10,000 executions, 50 active workflows	None
n8n Cloud Enterprise	Custom	Unlimited	None
Managed one-click (InstaPods/PikaPods)	$3-7/mo	Unlimited	None (platform-handled)
Raw VPS + Docker	$4-5/mo	Unlimited	Full (OS, Docker, n8n, SSL, backups, firewall — you)
Coolify on VPS	$5-8/mo	Unlimited	Partial (Coolify + VPS updates)
Elestio (managed)	~$17/mo	Unlimited	None (platform-handled)

When n8n Cloud Makes Sense Despite the Cost Premium

• Zero-ops teams — nobody wants to think about infrastructure at all.
• Low-volume automation — under 2,500 executions/month, Starter is genuinely fine and the premium buys real convenience.
• Enterprise compliance out-of-the-box — SSO, LDAP, audit logs without configuring anything yourself (Cloud Enterprise tier).
• Rapid prototyping — testing n8n before committing to infrastructure.

When Self-Hosting Wins

• Any meaningful, sustained automation volume — the cost delta compounds fast once you’re past a few thousand executions/month.
• Data residency / compliance requirements where you need to control exactly where data physically lives (healthcare, legal, regulated verticals).
• Teams with even light DevOps capacity — the “traditional self-hosting pain” (server setup, SSL, maintenance) is now largely absorbed by one-click platforms like InstaPods/PikaPods at $3-7/mo, closing much of the convenience gap that used to favor Cloud.

Open Questions

• No independently benchmarked comparison of n8n Cloud’s actual production throughput vs. a well-tuned self-hosted queue-mode setup — the 22→60.2 req/s figure comes from a single self-hosted benchmark video, not a head-to-head against Cloud.^[ambiguous]
• The ExpressTech cost source and the security-focused dev.to source are both third-party blogs (agency and individual author respectively), not n8n’s own cost/security documentation — the specific dollar figures and the compliance framing should be treated as directional, not official guidance. n8n’s own pricing page (linked from the official pricing-change blog) is the source to re-verify before making a purchasing decision.
• Multi-main high-availability setup requires Enterprise licensing on both self-hosted and Cloud — this article doesn’t have pricing for self-hosted Enterprise specifically (Cloud Enterprise is “custom”; self-hosted Enterprise pricing wasn’t surfaced by this research pass).

Related

• n8n vs Claude Code — When to Use Which — the higher-level tool-choice decision this article’s cost data supports
• Building n8n Workflows with Claude Code — the official MCP server (Q2 2026 update) ships identically across self-hosted and Cloud
• Zapier — Automation Platform and AI Agent Builder — Zapier’s cloud-only, task-based pricing is the contrast case to n8n’s self-hostable, execution-based model
• n8n Alternatives — Make and Pipedream (2026) — Make is cloud-only with no self-hosting option at all; Pipedream and n8n both offer developer-first self-hostable-or-cloud paths
• Managed Agents — Self-Hosted Sandboxes & MCP Tunnels — the parallel self-hosted-vs-managed trade-off in Claude’s own agent infrastructure

Try It

1. Model your real execution volume before picking a plan. Add up executions/month across every workflow you’d run (not just the busiest one) and check it against the Cloud Starter (2,500) and Pro (10,000) caps before committing.
2. Start with a managed one-click self-hosted platform (InstaPods, PikaPods) rather than raw VPS + Docker if you want self-hosting’s cost/execution benefits without taking on full sysadmin duty — it closes most of the “Cloud is more convenient” gap.
3. Don’t enable queue mode until you actually need it. It requires Postgres 13+ and Redis — added infrastructure complexity that only pays off once single-instance concurrency becomes a real bottleneck (roughly 50+ simultaneous workflow triggers as a rule of thumb).
4. If compliance is a named requirement (HIPAA, SOC 2, GDPR data residency), default to self-hosted and budget for the Business-tier-or-above SSO/LDAP features rather than assuming Community Edition alone satisfies an audit.