n8n Self-Host vs Cloud in 2026: The RevOps Decision Guide

Every RevOps team I talk to in 2026 is running n8n workflows for something mission-critical: lead routing, CRM enrichment, webhook processing from Clay or Apollo, AI agent orchestration. The question is no longer “should we use n8n” but “where do we run it?” The self-host vs cloud decision looks like a simple cost trade-off on the surface, but I’ve watched teams make expensive mistakes in both directions. This guide gives you the honest numbers and the actual decision framework.

The True Cost Breakdown

The n8n Cloud pricing page makes the math look simple. Starter is $20/month for 2,500 executions. Pro is $50/month for 10,000. Scale tiers go higher. But “executions” in n8n means full workflow runs, not individual node steps, which is actually favorable compared to how Make and Pabbly count operations.

On Make, every module step inside a scenario counts as an operation. A 12-step enrichment workflow that runs 1,000 times per month burns 12,000 operations. On the Make Core plan, you get 10,000 operations for $10.59/month, so that single workflow blows through your quota. At GTM scale running Clay-to-CRM enrichment pipelines or website visitor signals from Vector, you can easily hit 100k+ operations per month on Make and be looking at $59-299/month before you’ve built half your stack.

Pabbly Connect’s lifetime deal ($249-499 one-time) sounds like a steal. For simple linear workflows, it is. But Pabbly has a 30-minute minimum polling interval on most triggers, no native webhook server for instant triggers, and their “unlimited” plans throttle heavily on multi-step workflows. My clients who have migrated off Pabbly did so for one specific reason: their lead routing workflows were lagging 25-30 minutes behind form submissions. That is not a minor inconvenience when you are trying to hit a five-minute lead response time.

For self-hosted n8n, here is the honest cost breakdown I see with clients at Homegrown Growth Co.:

• VPS (Hetzner CX22 or DigitalOcean Basic): $6-20/month depending on region and specs
• Managed Postgres (Supabase free tier or Railway): $0-25/month
• Redis for queue mode (optional but recommended for >50k executions): $7-15/month
• Backups + object storage (S3 or Cloudflare R2): $1-5/month
• DevOps time: This is the one people forget. Upgrades, SSL cert renewals, monitoring, debugging failed workers. Averaged across a quarter, that runs 4-5 hours/month. At a $75 blended rate for someone on your RevOps team (not a dedicated DevOps hire), you are adding $300-375/month in real labor cost.

All-in: a properly maintained self-hosted n8n instance costs $350-440/month in total economic cost for a small GTM team. n8n Cloud Pro at $50/month with 10,000 executions is the obvious answer unless you have a dedicated infra person already on payroll (marginal cost near zero) or an execution volume that makes Cloud pricing a real budget line.

Where Each Option Actually Wins

Don’t choose self-host because you think it’s cheaper. Do the full math first, including your own time. Here’s how I frame it for clients.

n8n Self-Host vs Cloud vs Alternatives

Choose n8n Cloud if

• Your team has no dedicated DevOps or platform engineer
• You run under 50k executions per month across all workflows
• You need to move fast: onboarding takes minutes, not an afternoon

From $20/month Try n8n Cloud →

Choose n8n Self-Hosted if

• You have a DevOps person or platform engineer already on the team
• Data residency or compliance rules prevent third-party SaaS (HIPAA, SOC 2, EU data)
• You are running 500k+ executions per month and Cloud pricing becomes a real line item

Free software, ~$40-80/month infra Get n8n Self-Host →

Choose Make if

• Your workflows are mostly linear, low-step, and low-volume
• Your team prefers a no-code visual builder over n8n's more technical canvas
• You need pre-built templates for common marketing automation tasks fast

From $10.59/month (Core) Try Make →

Choose Pabbly Connect if

• You want a one-time payment and run simple, non-time-sensitive workflows
• Volume is high but workflow steps are minimal (under 5 steps per scenario)
• You are a solo operator or tiny team without budget for monthly SaaS

From $249 lifetime Try Pabbly →

How to Choose: The Framework I Actually Use

When I audit a client’s automation stack at Homegrown Growth Co., I ask four questions in order.

First: does anyone on this team maintain servers as part of their job description already? If yes, self-host is on the table. If no, stop there and use Cloud. The DevOps tax is real, and it falls on whoever is closest to the problem. In RevOps shops, that is usually the person who built the workflows, not someone who enjoys debugging Docker networking at 9pm.

Second: what is the current and projected cost per rep across your full stack? If you are already optimizing every line item, the $30/month delta between n8n Starter and self-host infra costs is noise. If you are running 300k+ executions on a workflow that processes Attio CRM enrichment jobs or lead routing at volume, then doing the math on Cloud’s execution tiers versus a $20 Hetzner box actually matters.

Third: do you have data residency requirements? EU-based teams with GDPR obligations or US healthcare-adjacent companies often cannot send workflow payloads through a vendor’s cloud. Self-host is not optional in those cases. It is the compliance decision. According to the n8n docs on self-hosting, you can run n8n completely air-gapped with outbound telemetry disabled.

Fourth: how often are you shipping new workflows? Cloud removes the “update the n8n version before adding this new node” friction entirely. When a client team is iterating weekly on new GTM automations, that friction compounds fast. I have seen teams stall for two or three days because a weekend upgrade broke their queue workers and they were afraid to roll back.

The Maintenance Burden Nobody Talks About

The hidden cost of self-hosting is not the server bill. It is compounding upgrade debt. n8n ships frequent releases, and running more than two major versions behind means community nodes start breaking, new AI agent nodes are unavailable, and you cannot reproduce issues against the current docs. I have seen self-hosted instances running n8n 0.214 when the team was actively trying to implement features from the 1.x documentation. That is a painful place to be.

The upgrade path for a properly configured self-hosted instance (Docker Compose or Kubernetes, external Postgres, Redis in queue mode) is not a docker pull and restart. You need to test against existing workflows, check for breaking changes in the changelog, and have a rollback plan. Teams that skip this end up with a brittle instance they are afraid to touch. And then they stop touching it. And then it falls further behind.

# Minimum production-grade docker-compose.yml
services:
  n8n:
    image: n8nio/n8n:latest
    environment:
      - DB_TYPE=postgresdb
      - DB_POSTGRESDB_HOST=your-managed-db-host
      - EXECUTIONS_MODE=queue
      - QUEUE_BULL_REDIS_HOST=your-redis-host
      - N8N_ENCRYPTION_KEY=${ENCRYPTION_KEY}
    volumes:
      - n8n_data:/home/node/.n8n
  worker:
    image: n8nio/n8n:latest
    command: worker
    environment:
      - EXECUTIONS_MODE=queue
      - QUEUE_BULL_REDIS_HOST=your-redis-host

The Bottom Line on n8n Deployment in 2026

The self-host vs cloud decision is really a staffing and volume question, not a software question. Cloud is the right default for the vast majority of RevOps teams, especially those building on top of n8n as their primary GTM automation layer. Self-host makes sense the moment you have free DevOps capacity, a compliance requirement that rules out SaaS, or execution volumes above 500k per month where Cloud pricing tiers create a real budget conversation. Make remains viable for simpler, lower-volume scenarios, and Pabbly earns a look for solo operators who want to cap variable costs. For teams running serious GTM infrastructure, though, n8n Cloud’s execution-based model is genuinely more predictable than Make’s operation counting at scale.