N8n replaces Zapier the moment you own a server and two spare hours

In July 2026, Zapier charges $29.99/month for 750 tasks — about four cents per automated action. Meanwhile, n8n ships version 2.32.5 (stable) and 2.33.0 (pre-release, July 28), running a pricing model that makes every SaaS competitor look expensive: unlimited executions, self-hosted, zero subscription. The LTS track, 1.123.67 from July 22, keeps the enterprise channel stable. The project, founded in Berlin in 2019 by Jan Oberhauser, has passed 198,000 GitHub stars and shows no sign of slowing down.

Zapier bills per task — every step inside a workflow is a chargeable unit. N8n bills per execution: the entire workflow run, start to finish. A 50-step workflow that fires 1,000 times per month costs 1,000 executions on n8n. On Zapier, it costs 50,000 tasks. The annual gap easily exceeds $8,000.

What n8n is — and what it isn’t

N8n is a fair-code automation platform under the Sustainable Use License: the source is public, you can use it, modify it, and redistribute it for free, as long as you don’t resell n8n as a managed service. For internal use — syncing your CRM, routing emails, orchestrating notifications — the license is free and uncapped.

The interface is a visual canvas where each node represents an action: webhook trigger, API call, JSON transformation, email send, SQL query. Over 400 built-in integrations cover the mainstream — GitHub, Slack, Notion, Google Sheets, PostgreSQL, Telegram, Home Assistant — and a generic HTTP Request node fills every gap.

What sets n8n apart from SaaS alternatives:

• Unlimited executions on self-hosted. A $5/month VPS runs as many workflows as you want, no metering.
• Native code engine. JavaScript and Python nodes execute arbitrary code inside the workflow — no sandboxed runtime or runtime limits like Zapier’s 30-second cap.
• Cron and webhook triggers. Precise scheduling and inbound HTTP endpoints, no extra charge.
• AI nodes. The AI Agent node connects n8n to OpenAI, Anthropic, Ollama, or any API-compatible model — your workflows become autonomous agents.
• Git-native. Workflows export as JSON, version in any repo, share across instances.

N8n is not a SaaS, and that’s by design: no fair-use cap at 10,000 executions, no step-count limit per workflow, no premium apps that vanish when you downgrade.

The bill: a table that does the talking

Take a realistic 8-step workflow — new HubSpot contact → Clearbit enrichment → Salesforce update → Asana task → Slack notification → Google Sheets log → confirmation email → calendar reminder — running 1,000 times per month.

Same workflow, same volume. Zapier costs 3–7× more than Make, and 50–100× more than self-hosted n8n. And the gap widens with complexity: a 16-step workflow doubles Zapier’s bill but changes nothing for n8n.

Scale it to 100,000 records per month with a 5-step workflow:

The annual gap between Zapier and self-hosted n8n exceeds $6,900. For a small business or freelancer, that’s a line item that becomes pure margin.

The docker-compose that boots in five minutes

Deploying n8n is a single file. Here’s the minimal production config behind Traefik:

version: "3.8"

services:
  n8n:
    image: docker.n8n.io/n8nio/n8n
    container_name: n8n
    restart: always
    environment:
      - N8N_HOST=n8n.your-domain.com
      - N8N_PORT=5678
      - N8N_PROTOCOL=https
      - WEBHOOK_URL=https://n8n.your-domain.com
      - N8N_ENCRYPTION_KEY=your-encryption-key
      - DB_TYPE=postgresdb
      - DB_POSTGRESDB_HOST=postgres
      - DB_POSTGRESDB_DATABASE=n8n
      - DB_POSTGRESDB_USER=n8n
      - DB_POSTGRESDB_PASSWORD=${DB_PASSWORD}
    volumes:
      - n8n_data:/home/node/.n8n
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.n8n.rule=Host(`n8n.your-domain.com`)"
      - "traefik.http.services.n8n.loadbalancer.server.port=5678"

  postgres:
    image: postgres:16
    container_name: n8n_db
    restart: always
    environment:
      - POSTGRES_DB=n8n
      - POSTGRES_USER=n8n
      - POSTGRES_PASSWORD=${DB_PASSWORD}
    volumes:
      - n8n_db:/var/lib/postgresql/data

volumes:
  n8n_data:
  n8n_db:

A few production caveats:

• N8N_ENCRYPTION_KEY is mandatory — generate one with openssl rand -hex 32 and store it somewhere safe. It encrypts every credential.
• PostgreSQL is strongly recommended above a dozen workflows; the default SQLite backend becomes a write bottleneck beyond that.
• Never expose n8n directly without a reverse proxy — the service has no built-in TLS. Place it behind Nginx Proxy Manager, Traefik, or Caddy.

Startup takes under two minutes on a 2 GB RAM VPS. The web UI is available on port 5678.

AI, MCP, and what’s landing in 2026

N8n invested heavily in AI during 2025–2026. The AI Agent node lets you build workflows that call an LLM, parse its response, and loop over tools — intent detection, email classification, structured data extraction from raw text. Version 2.33.0 (pre-release, July 28, 2026) improves Telegram and Slack channel setup for agents and stabilizes MCP (Model Context Protocol) connections.

LangChain and LlamaIndex nodes are available through the community library, and the Ollama node runs local models with zero cloud dependency. A typical workflow: inbound webhook → local LLM classification → conditional routing → API action — entirely self-hosted, no external calls.

Beyond AI, the community node ecosystem has grown into a serious asset. Nodes for ntfy, Paperless-ngx, Uptime Kuma, and Home Assistant bridge n8n to the rest of your self-hosted stack — the kind of integrations that Zapier and Make simply don’t ship because the user base for a self-hosted document scanner or DNS filter is too niche for a SaaS product team to prioritize. On n8n, if a node doesn’t exist, you build it once and reuse it forever.

Verdict: who should switch, who should stay

Switch to n8n if:

• You already run a server (VPS, NAS, mini PC) 24/7 — the marginal cost is zero.
• Your workflows exceed 500 tasks per month — the switch pays for itself within the first month.
• You need code nodes (Python, JavaScript) that Zapier limits or charges extra for.
• You want to connect self-hosted services (Home Assistant, Paperless-ngx, ntfy) that SaaS platforms ignore.
• Data sovereignty matters: your credentials and execution logs stay on your own disk.

Stick with Zapier if:

• You handle fewer than 100 tasks per month with simple 2–3 step workflows — the free tier works.
• Your critical integration has no native n8n node and you’d rather not write a manual HTTP Request.
• Nobody on your team can maintain even a minimal server — human time costs more than the subscription.

For everyone else — the people already running Immich, Vaultwarden, or Jellyfin — adding n8n to the docker-compose takes ten minutes and frees up a monthly budget you can redeploy immediately. As of July 29, 2026, the question isn’t whether n8n is mature enough. It’s why you’re still paying Zapier.

References

• n8n GitHub — Releases — versions 2.33.0, 2.32.5, 1.123.67
• n8n Documentation — Sustainable Use License
• Automation Atlas — Zapier vs Make vs n8n (2026)
• AICentralResources — Zapier vs Make vs n8n Pricing 2026
• selfh.st — weekly self-hosted newsletter
• noted.lol — self-hosted software reviews