n8n is the workflow tool we run for any client past 10,000 tasks/month. The reason isn't the integration catalog — Zapier wins that. It's that the bill stops growing with the work. Self-hosted n8n on a $5 VPS handles workloads Zapier would charge $400/month for.
n8n (pronounced "n-eight-n", short for "nodemation") is an open-source workflow automation platform. You build automations by chaining together visual nodes — each node is a connector to a service (Slack, Gmail, HubSpot, OpenAI, your own database) or a logic step (if/else, loop, wait). It is in the same category as Zapier and Make, but with three differences that matter: it is open-source, it is code-friendly, and it is self-hostable. For the broader pattern view of when to reach for n8n vs Make vs Zapier — and the 5 patterns that run in production — see our guide on what AI automation actually is.
Who n8n is for
- Technical operators who want to drop into JavaScript or Python when the visual nodes don't cover their case.
- Teams running AI-heavy workflows where Zapier's per-task pricing would explode.
- Companies with strict data residency or compliance needs (HIPAA, EU customer data) who self-host.
- Agencies and consultants who build automation for clients and want to own the IP.
- Builders who want to wire up internal tools with the same engine that drives their customer-facing automations.
It's not the right fit for: a solo founder who has never seen JSON before, or a marketer who just wants to connect two SaaS tools and walk away.
What you can actually build with it
AI agent workflows
n8n has first-class nodes for OpenAI, Anthropic, and Google AI. You can build an agent that watches an inbox, drafts replies, classifies leads, or runs multi-step research — all visually, with custom code where you need it.
CRM and marketing automation
Lead enrichment, scoring, routing, follow-up sequencing, attribution reporting. Same shape as what HubSpot Workflows or GHL Automation does, but vendor-neutral and cheaper at scale.
Data pipelines
Pull from one system, transform with code, push to another. Daily syncs, hourly refreshes, on-demand triggers. Functions as a poor-man's ETL when full Airbyte / Fivetran is overkill.
Internal tooling
Slash commands, approval workflows, on-call escalation, custom dashboards via webhooks. Replaces a category of glue work that used to require an internal eng team.
How n8n is different from Zapier
Three structural differences separate n8n from Zapier and they all favor heavier workflows. n8n cloud charges per workflow execution regardless of step count, where Zapier charges per task — for any workflow over 5 steps, n8n runs 5–20x cheaper at the same volume. n8n treats code as first-class with full Python/JavaScript and npm package access, where Zapier code steps are second-class and limited. n8n self-hosts in Docker on infrastructure you own, where Zapier is cloud-only — the deciding factor whenever data cannot leave the environment for compliance or vendor-risk reasons.
- Pricing model. Zapier charges per task (each step in a workflow). n8n cloud charges per workflow execution, regardless of step count. For complex workflows, n8n is 5–20x cheaper.
- Code support. Both have "code steps", but n8n's are first-class. You can write JavaScript or Python with full access to npm packages, install custom dependencies, and treat the platform like a backend.
- Self-hosting. n8n runs in Docker on your own server. Zapier is cloud-only. If your data can't leave your environment, this is the only choice.
Other meaningful differences: n8n has fewer pre-built integrations than Zapier (~400 vs Zapier's ~6,000), but you can build any HTTP-based integration in 10 minutes. n8n's UI is denser and more developer-coded; Zapier's is friendlier for non-technical users.
How n8n is different from Make (Integromat)
Make and n8n are closer cousins than n8n and Zapier — both have visual flow editors with branching and parallelism — but they split on four dimensions that matter for procurement. n8n is open-source and self-hostable; Make is cloud-only. n8n has a much stronger story for code-heavy workflows; Make has nicer error visualization and has been more reliable in our usage at high concurrency. n8n cloud prices per execution; Make prices per "operation" (somewhere between Zapier's task and n8n's execution model). Most teams pick n8n over Make when they want to self-host or when they have engineers, and pick Make when they want a slightly more polished cloud experience for non-technical operators.
- n8n is open-source and self-hostable. Make is cloud-only.
- n8n has a much stronger story for code-heavy workflows.
- Make has nicer error visualization and has been more reliable in our usage at high concurrency.
- Make pricing is per "operation" (somewhere between Zapier's task and n8n's execution); n8n cloud pricing is per execution.
Most teams pick n8n over Make when they want to self-host or when they have engineers; pick Make when they want a slightly more polished cloud experience for non-technical operators.
Self-hosting vs cloud
| Dimension | Self-hosted | n8n Cloud |
|---|---|---|
| Cost | $5–20/mo VPS only | $20–50/mo + tier |
| Executions | Unlimited | Metered per tier |
| All features | Yes — full open-source build | Tier-gated (SSO, audit logs above Pro) |
| Updates / backups | You handle (1 hr/mo if set up well) | Automatic |
| Scaling / HA | Your problem — needs DevOps | Managed |
| Data residency | Anywhere Docker runs (your infra) | EU / US (n8n's infra) |
| Best for | Engineers, high volume, compliance | Zero-ops, moderate volume |
Common n8n architectures we ship
- AI inbox triage: webhook → classify with Claude → route to GHL → notify Slack on edge cases.
- Lead enrichment: form submission → enrich with Apollo / Clearbit → score → push to CRM with smart owner assignment.
- Daily ops report: cron → query 5 systems → summarize with LLM → post to Slack and email.
- Calendly + AI prep: booking webhook → research attendee → draft brief → email to host 30 min before call.
- GHL bridge: any system → n8n → GHL via API. We use n8n constantly to extend GHL beyond what its native automation supports.
Limitations to know about
- Fewer pre-built integrations than Zapier. Workaround: HTTP node + 5 minutes of API docs covers anything with an API.
- No official mobile app. Workflows that need to be triggered or monitored from phones use webhooks + a dashboard or Slack.
- The visual editor gets unwieldy past ~50 nodes. Modular sub-workflows are the answer; new operators don't always discover this fast enough.
- Self-hosting requires real DevOps if you want HA, multi-region, or audit-grade observability.
How to evaluate if n8n is right for you
n8n is the right pick when at least two of three conditions hold: you have someone on the team who can read code (even occasionally), you run more than 10,000 task-equivalents per month (where execution-pricing economics start to dominate task-pricing), or you need to self-host for compliance or platform-ownership reasons. If two of three are yes, n8n almost always wins on cost and capability ceiling. If all three are no, you'll be happier on Zapier — the lower friction is worth the higher per-task cost at low volumes.
- Do you have anyone on the team who can read code, even if they don't write it daily? (No → use Zapier or Make.)
- Are you running, or about to run, more than 10,000 task-equivalents per month? (Yes → n8n economics start to dominate.)
- Do you need to self-host for compliance or want to own the platform? (Yes → n8n is one of the only realistic options.)
If you answered yes to two of three, n8n is probably the right pick. If you answered no to all three, you're happier on Zapier.