n8n

I have been running n8n self-hosted on a $6/month VPS since January 2025. It processes about 15,000 workflow executions a month — email parsing, Slack notifications, database syncs, AI-powered content classification — and my total monthly cost is still that same six dollars. My previous Zapier bill for roughly the same workload was $130/month.

That price difference — $6 versus $130 — is the entire n8n story in one sentence. But it is also not the whole story, because n8n asks something of you that Zapier and Make do not: a willingness to get your hands dirty.

I tested n8n alongside Make.com and Zapier for two weeks of dedicated comparison testing, on top of the year-plus I have been running it in production. Here is what I found: n8n is the best automation platform for people who can spell "Docker" without googling it. For everyone else, Make or Zapier might still be the right call.

What n8n actually is

n8n is an open-source workflow automation platform. You connect apps and services through a visual node editor — drag a node onto the canvas, configure it, connect it to the next node, and you have an automation. It runs on your own server (self-hosted, free) or on n8n's cloud (paid, managed).

The core product has three parts:

The workflow editor. A visual canvas where you build automations by connecting nodes. Each node represents an action: "read emails from Gmail," "filter by subject line," "create task in Notion," "send Slack message." The editor shows data flowing between nodes so you can trace what happens at each step. It is similar to Make.com's visual editor but less polished — think "engineer's tool" versus "designer's tool."

The node library. Over 400 pre-built integrations covering most major services: Google (Gmail, Drive, Sheets, Calendar), Microsoft (Outlook, Teams, Excel), Slack, Discord, Notion, Airtable, GitHub, GitLab, Stripe, OpenAI, and many more. Each node is essentially a wrapper around an API with a configuration UI. If a service is not in the library, you can use HTTP Request nodes to call any REST API, or write custom code nodes in JavaScript or Python.

The execution engine. This is where n8n differs most from Zapier. Zapier polls for triggers on a schedule (every 1-15 minutes depending on your plan). n8n can use webhooks for instant triggers, supports complex branching and error handling, and executes workflows with full context about what happened at each step. The result is faster, more reliable automations — but also more complexity to configure.

The AI-native features added in late 2024 are worth mentioning separately. n8n now includes nodes for LLM chat (OpenAI, Anthropic, local models via Ollama), text classification, embeddings generation, and vector store operations (Pinecone, Qdrant, Weaviate). You can build workflows that process documents through AI, classify support tickets, generate summaries, or trigger actions based on AI analysis — all within the same visual editor.

Self-hosted: the real reason to choose n8n

The self-hosted option is the feature that makes every other comparison almost irrelevant. n8n's open-source version is fully functional — no feature gates, no usage limits, no "contact sales for this node" nonsense. You deploy it on your own server (a single Docker command on any Linux machine) and you get the entire platform for free, forever.

I run it on a Hetzner VPS with 2 vCPUs and 4GB RAM ($6/month). That handles 15,000 monthly executions with room to spare. For heavier workloads, you would scale up the server, but the math stays favorable: a $20/month VPS handles roughly 50,000-100,000 executions per month. Zapier charges $130/month for 10,000 tasks on their Team plan and the pricing climbs steeply from there.

The catch: you are responsible for everything. Server maintenance, SSL certificates, database backups, n8n version updates, monitoring, security patches. This is not hard if you have basic DevOps skills — Docker Compose handles most of it, and upgrade paths are well-documented — but it is real work. Budget an hour a month for maintenance. If your time is worth more than the price difference between self-hosted and cloud, pay for Cloud.

Also worth noting: some enterprise features (SSO, audit logs, advanced permissions) are Cloud-only. The open-source version has basic user management but not the compliance features larger organizations need. If you need SOC 2 or SAML, you are on Cloud Pro.

The self-hosted installation itself is straightforward. On a fresh Ubuntu server:

docker run -d --name n8n -p 5678:5678 \
  -v n8n_data:/home/node/.n8n \
  -e N8N_SECURE_COOKIE=false \
  n8nio/n8n

That is it. Visit http://your-server:5678 and you are in the editor. Production deployments want more configuration (SSL, authentication, persistent storage, database backend), but the getting-started experience is genuinely that simple.

AI workflows: the 2026 differentiator

n8n's AI nodes are not as sophisticated as dedicated AI orchestration platforms like LangChain or Relevance AI. But for the common automation use cases — classify this, summarize that, extract structured data from these — they work well and are deeply integrated into the workflow editor.

I built three AI-powered workflows during testing:

Support ticket classifier. Incoming emails hit an n8n webhook, get classified by GPT-4o-mini into categories (billing, technical, feature request, spam), and are routed to the right Slack channel or support queue. Classification accuracy is about 85-90%, which is good enough to save a human 15 minutes of triage per day. Cost: about $0.30/month in OpenAI API fees.

Content moderation pipeline. User-submitted content goes through a chain: text classification node flags potentially problematic content, a second LLM node rewrites flagged content with guidelines, and a human review step is created in Notion if confidence is below 90%. The chaining is the impressive part — each node passes structured data to the next, and you can see what the AI decided at each step.

Meeting notes processor. A workflow that watches a Google Drive folder for new meeting transcripts, summarizes them with Claude, extracts action items into a structured format, creates tasks in Notion, and emails participants a summary. This replaced a manual process that took me 20 minutes per meeting.

The AI nodes are not going to replace a dedicated AI pipeline for complex use cases. You cannot fine-tune models, run evaluations, or manage prompt versioning in n8n. But for "add AI to my existing automation" — which is what most people actually need — the integration is seamless and works well.

Where n8n wins

Price at scale. This cannot be overstated. Zapier and Make charge per task or per operation, and those costs compound fast. A workflow that processes 50,000 items per month costs $300-800 on Zapier (Team or Business plan). The same workflow on n8n costs whatever your server costs — $6-30/month. Over a year, that is thousands of dollars saved. Over three years, it is tens of thousands.

Full control and transparency. Because n8n is open-source, you can audit the code, modify nodes, and trust that your data is not being processed on servers you do not control. For regulated industries (finance, healthcare, legal), this matters enormously. Zapier and Make process your data on their infrastructure; n8n self-hosted processes it on yours.

Developer-friendly extensibility. Custom code nodes let you write JavaScript or Python inline within any workflow. Need to transform data in a way no built-in node handles? Write 10 lines of code. Need to call an API that is not in the node library? HTTP Request node plus a code node for parsing. This flexibility means you almost never hit a hard limit on what n8n can do — you just write the missing piece.

Webhook-native architecture. n8n workflows can be triggered instantly via webhooks rather than polling on a schedule. This means faster automations and lower API usage (you are not checking "any new emails?" every 5 minutes). Zapier recently added webhook triggers on higher-tier plans; n8n has had them from day one on every tier including self-hosted.

Community and ecosystem. n8n has an active community building and sharing workflow templates. The node library grows through community contributions. If you need to automate something obscure, there is probably a community node for it. The Discord is responsive and the documentation is good — better than Make's and competitive with Zapier's.

Where n8n loses

The UI is functional, not beautiful. n8n's editor works fine but it shows its developer roots. Nodes are boxes with text labels. The data flow visualization is basic. Error messages are sometimes cryptic. Compare to Make.com, which shows animated data flowing between beautifully designed nodes and makes complex data transformations feel intuitive. If aesthetics and approachability matter to your team, Make wins this category decisively.

Fewer integrations than Zapier. Zapier has 7,000+ integrations. n8n has 400+. That is a meaningful gap. If your automation depends on niche SaaS tools, check the n8n node library before committing. You can always use HTTP Request nodes to bridge the gap, but that requires reading API documentation and writing auth headers — exactly the kind of work automation tools are supposed to eliminate.

Setup friction. Self-hosting requires a server, Docker knowledge, and ongoing maintenance. Cloud removes the maintenance but costs more (€24/month for Starter). Either way, the initial setup takes longer than signing up for Zapier, which you can do in 30 seconds with a Google account. This friction is a real barrier for non-technical users.

Error handling could be better. n8n has error workflows (trigger an action when something fails) and retry logic, but the debugging experience is weaker than Zapier's. When a workflow fails, Zapier shows you exactly which step broke, what the input was, and often suggests a fix. n8n shows you an error message and execution data and leaves you to figure it out. For complex workflows with many steps, this can be frustrating.

Cloud pricing is in euros. A minor point but worth noting: n8n Cloud bills in euros. At current exchange rates (May 2026, roughly €1 = $1.08), the €24 Starter plan is about $26/month. Not a dealbreaker but worth factoring into comparisons with dollar-denominated competitors.

Pricing: the honest comparison

Here is what you actually pay at different usage levels:

Light usage (2,000-5,000 executions/month):

• n8n Self-Hosted: $6/month (VPS) or free (run on an old laptop)
• n8n Cloud Starter: €24/month (~$26)
• Zapier Starter: $30/month (750 tasks — probably not enough)
• Zapier Professional: $74/month (2,000 tasks)
• Make.com Core: $10/month (10,000 operations)

At light usage, Make.com is actually the cheapest managed option at $10/month. n8n self-hosted beats it on price if you already have a server, but Make's visual editor is better for simple workflows.

Medium usage (10,000-25,000 executions/month):

• n8n Self-Hosted: $12-20/month (better VPS)
• n8n Cloud Pro: €60/month (~$65)
• Zapier Team: $130/month (10,000 tasks)
• Zapier Business: $337/month (25,000 tasks)
• Make.com Teams: $35/month (40,000 operations)

This is where n8n self-hosted pulls away. At 25,000 executions, n8n costs $20/month (server) versus $337/month on Zapier Business. That is $3,800/year saved.

Heavy usage (50,000+ executions/month):

• n8n Self-Hosted: $30-60/month (dedicated server)
• n8n Cloud: Enterprise pricing (negotiated)
• Zapier: $800+/month (and you are probably talking to sales)
• Make.com Enterprise: Custom pricing

At scale, the self-hosted math is overwhelming. A $50/month server handles 100,000+ monthly executions. Zapier and Make cannot touch that on price.

The one asterisk: these are execution counts, not task counts. Zapier counts each action as a task (a 5-step workflow run once = 5 tasks). n8n counts the entire workflow run as one execution regardless of how many steps it has. So the price gap is often even larger than it appears.

Who should use n8n

Buy n8n if:

• You have basic DevOps skills and are comfortable running a server
• Your automation volume is high enough that per-task pricing hurts (10,000+ tasks/month)
• You need to keep data on your own infrastructure for compliance reasons
• You want to build automations that include custom code or AI processing
• You are price-sensitive and prefer predictable infrastructure costs over variable SaaS pricing
• You are a developer or work on a technical team

Skip n8n if:

• You want the smoothest possible setup with zero maintenance
• Your team is non-technical and needs a drag-and-drop tool that "just works"
• You depend on niche SaaS apps that n8n does not have nodes for (check the library first)
• You want a beautiful, polished UI that non-technical stakeholders can navigate
• You are doing fewer than 2,000 automations per month (Zapier Starter or Make Core might be simpler)

The verdict

n8n is the automation platform I recommend to anyone who knows what a VPS is. The combination of open-source transparency, unlimited free self-hosting, and a genuinely capable workflow editor makes it the best value in the automation category by a wide margin.

It is not for everyone. The setup friction is real. The UI, while functional, will not win design awards. And the integration library, while solid, does not match Zapier's breadth. If you need the easiest possible experience and money is less of a concern, Make or Zapier are still fine choices.

But if you are building automations that will run for years, processing thousands of events per month, the math tilts hard toward n8n. A $6/month VPS versus a $300/month Zapier bill is not a close comparison. And the open-source model means you are not betting your business on a company that might get acquired, change pricing, or shut down.

I have been running n8n in production for over a year. It has never gone down, never lost data, and never surprised me with a bill. That is more than I can say for any SaaS automation tool I have used.

n8n review updated May 2026. Testing conducted on self-hosted (v1.80+) and Cloud Pro plans, processing ~15,000 real workflow executions per month since January 2025. Comparisons made against Zapier Professional/Team and Make.com Core/Teams. AI workflows tested with OpenAI GPT-4o-mini and Claude 3.5 Sonnet. Affiliate disclosure: LaunchToolsAI may earn a commission if you subscribe through our links.