TL;DR: n8n (pronounced "n-eight-n," short for "nodemation") is a visual workflow automation platform you can either run free on your own server or pay for as managed cloud hosting. It has 1,700+ pre-built integrations and 70+ native AI/LangChain nodes. Compared with Zapier or Make, n8n is more flexible, friendlier to developers, and dramatically cheaper at scale — but harder for non-technical users on day one. This page explains what it is, who built it, what you can build with it, and how it compares with the alternatives.
What n8n actually is
n8n is a workflow automation platform — software that lets you connect different apps, APIs, databases, and AI models so they can pass data between each other automatically, without you having to write the glue code yourself. If you've heard of Zapier, Make.com, or Microsoft Power Automate, n8n sits in the same category but with a different philosophy: it's source-available (you can read and modify the code), it can run on your own infrastructure for free, and it's built for developers as a first audience and non-developers as a secondary one.
You build workflows visually: drag boxes (called "nodes") onto a canvas, draw lines between them, and configure each box with the credentials and parameters it needs. Each node does one specific thing — "fetch new rows from Postgres," "send a Slack message," "ask an LLM to summarise this text," "store this embedding in a vector database." Connect them together and you have an automated pipeline that runs on a schedule, on demand, or in response to a webhook.
The product launched in 2019, is built in TypeScript on Node.js, and has crossed 200,000 stars on GitHub as of early 2026. It's developed and supported by n8n GmbH, a Berlin-based company that's raised a Series B from Highland Europe.
Who built n8n and why
n8n was founded by Jan Oberhauser, who'd previously worked in visual effects engineering. The story Jan tells is one many automation tool founders share: he kept finding himself writing Python or JavaScript glue scripts to move data between SaaS tools. Each script was brittle. Each one cost engineering time. The existing visual tools at the time — Zapier was already established, Integromat (which became Make) was growing — felt either too expensive at scale or too limited for non-trivial logic. So he built his own.
The fair-code positioning was a deliberate response to two trends in 2019. First, open-source projects were getting acquired or relicensed in ways that hurt users (the MongoDB SSPL drama, Elastic vs AWS). Second, single-vendor SaaS automation was getting expensive once your workflows got serious. Fair-code aimed to be the middle ground: source-available, free to self-host, free to modify, but with explicit restrictions that prevent cloud providers from reselling n8n as their own managed service.
How n8n differs from Zapier, Make, and Power Automate
The headline differences:
| Feature | n8n | Zapier | Make.com | Power Automate |
|---|---|---|---|---|
| Self-hostable | Yes (free) | No | No | No (cloud only) |
| Source-available | Yes (fair-code) | No | No | No |
| Pricing unit | Workflow execution | Per task / step | Per operation / credit | Per flow run + premium connectors |
| Integrations (native) | 1,700+ | 7,000+ | 3,000+ | 1,000+ (Microsoft-heavy) |
| Native AI / LangChain nodes | 70+ | ~10 | ~15 | ~20 (Azure AI focused) |
| Built-in JavaScript / Python code nodes | Yes | Limited (Code by Zapier, $) | Limited | No |
| Free tier | Self-host free, 14-day cloud trial | 100 tasks/month | 1,000 ops/month | Included in M365 plans |
| Starting paid price | €24/month (Cloud) | $19.99/month | $10.59/month | $15/user/month standalone |
The short interpretation: Zapier wins on breadth and ease of use, n8n wins on depth and price-at-scale, Make wins on visual polish and free tier, Power Automate wins inside Microsoft-first organizations. For most technical teams making this comparison in 2026, n8n is the new default — but Zapier remains the best on-ramp for non-technical operators.
What you can build with n8n: real workflows
A few concrete examples of what people actually build:
AI customer support agent (RAG pipeline)
This is the killer 2026 use case. The workflow: a webhook receives a customer support email → an OpenAI Embeddings node generates a vector for the email content → a Pinecone or Supabase pgvector node retrieves the 5 most relevant knowledge base chunks → a prompt template node assembles the question + context → an Anthropic Claude or OpenAI chat node generates an answer with citations → a conditional node checks confidence → if confident, post directly to Zendesk; if not, route to a human agent in Slack. The entire pipeline is built visually in 20-30 minutes.
Lead enrichment + CRM sync
Form submission → check duplicate in HubSpot → if new, call Clearbit and Apollo APIs to enrich → calculate lead score with an LLM-based prompt → if hot lead, create Salesforce opportunity + Slack DM the AE; if cold lead, add to nurture sequence in HubSpot. This is the 12-step workflow every RevOps team in B2B SaaS rebuilds.
Multi-source content publishing
New row in Airtable (your content calendar) → fetch the linked Google Doc → convert to Markdown → use GPT to generate platform-specific variants for LinkedIn, Twitter, and Mastodon → schedule posts via Buffer → log analytics back to the Airtable row.
Internal admin tools
Slack slash command (/lookup user) → query Postgres → format result → reply with a structured Slack message. Or: a daily scheduled query against your data warehouse → assemble a key-metrics report → email to leadership. The kind of internal tooling that used to require building a custom admin app now takes 30 minutes in n8n.
DevOps and incident response
PagerDuty incident webhook → fetch incident details → look up service owner in your service catalog API → DM owner in Slack with rich context → update status page → wait for ack → escalate if unacked after 10 minutes. All inside your VPC if you self-host.
Core concepts every n8n buyer should know
Nodes
Each box in a workflow is a node. Each node does one thing — fetch data, transform data, send data somewhere, decide what to do next, or wait. n8n ships with 1,700+ pre-built nodes covering most major SaaS apps, plus generic "protocol" nodes like HTTP Request that let you talk to anything with an API.
Executions
One execution = one complete run of a workflow, from the trigger node to the last reachable node. This is the unit n8n bills on. A workflow with 200 nodes processing 100 items in one run still counts as one execution. This pricing model is the single biggest reason engineering teams migrate to n8n from Zapier — your bill doesn't grow when your workflows grow.
Expressions
The way you reference data between nodes. They look like {{ $node["HTTP Request"].json["email"] }} or {{ $now.toFormat('yyyy-MM-dd') }}. Once you internalise expression syntax (typically a 4-hour learning curve for someone with no coding background), n8n becomes radically more flexible than Zapier.
Workflows, sub-workflows, and the canvas
Workflows can call other workflows via the Execute Workflow node — useful for sharing common logic. A typical mature n8n deployment has 30-50 workflows, with 5-10 reusable sub-workflows. Note that each sub-workflow call counts as a separate execution against your monthly cap.
The Code node
When the visual builder hits a wall, drop in a Code node. JavaScript by default, Python optionally. This escape hatch is what differentiates n8n from no-code tools — you're never blocked by a missing built-in transformation. The Code node makes n8n a viable platform even for complex transformations that would require 5+ Zapier "Formatter" actions chained together.
Credentials
API keys, OAuth tokens, and database connection strings are stored as credentials, isolated per environment, and shared across workflows that need them. Self-hosted Enterprise plans support external secrets managers (HashiCorp Vault, AWS Secrets Manager, Azure Key Vault) for organizations that have a "no secrets in app databases" policy.
How n8n makes money (and the fair-code question)
n8n is venture-backed and monetizes three ways: (1) n8n Cloud subscriptions (managed hosting), (2) self-hosted Enterprise licenses (SSO, RBAC, audit logs, external secrets), and (3) embedded n8n (for SaaS companies that want n8n inside their own product). The Community Edition is free forever — there is no plan to gate features behind a paid tier.
The "fair-code" license is worth understanding. The Sustainable Use License grants you the right to use, modify, and run n8n within your organization for any purpose, including commercially. What it restricts: you cannot offer n8n as a hosted service for third parties. So you can run n8n for your own company, even your own customers — but you cannot launch "Hosted n8n by MyCompany" as a paid product. For 99% of buyers this restriction is irrelevant. For the 1% who care, Activepieces (MIT licensed) is the closest true-open-source alternative.
Who should use n8n (and who shouldn't)
You should use n8n if…
- You're a technical team or have at least one engineer who's comfortable with APIs and JSON.
- You're already running into Zapier or Make pricing pain (€100+/month) and the next plan tier feels like extortion.
- You're building AI agents, RAG pipelines, or any workflow that calls multiple LLMs and vector stores.
- You need to run automations inside your own infrastructure for compliance reasons (healthcare, finance, government).
- Your stack is heavy on databases (Postgres, Mongo, Snowflake) — n8n's native database nodes are vastly better than Zapier's paywalled equivalents.
- You want to read and modify the source code of your automation platform.
You probably shouldn't use n8n if…
- You're a non-technical solo operator and the workflows you need are 2-step "send a Slack message when X happens" patterns. Stick with Zapier.
- You need 1-click templates for every imaginable use case (Zapier's marketplace is 4x bigger).
- Your business is a Microsoft 365 monoculture and Power Automate is bundled into your existing licensing.
- You require strict OSI-approved open source as a procurement policy. Look at Activepieces or Apache Airflow instead.
- You want HIPAA-compliant managed cloud hosting today (n8n Cloud doesn't BAA yet — self-hosted Enterprise is the route).
How to get started with n8n in 30 minutes
Two recommended paths:
Path A — n8n Cloud trial (5 minute setup)
- Sign up at n8n.io for the 14-day Cloud trial (no credit card required).
- Open the editor at your.n8n.cloud subdomain.
- Click "Start with a template" — pick a 3-step starter like "Slack notification on new HubSpot contact."
- Connect credentials when prompted. Test. Activate.
Path B — Self-hosted Community Edition (15 minute setup)
- SSH into a Linux server (DigitalOcean droplet, Hetzner VPS, AWS EC2 — anything with 2GB+ RAM).
- Install Docker.
- Run
docker run -it --rm -p 5678:5678 docker.n8n.io/n8nio/n8nfor a quick test. For production use, follow the official self-hosting guide with Postgres and Caddy/nginx. - Browse to
http://server-ip:5678and create your first admin account.
From either starting point, the typical learning trajectory: first workflow in 30 minutes, comfortable with expressions and Code nodes in two weeks, building production AI agents in a month.
The n8n ecosystem
n8n's community is genuinely active. The official community forum sees hundreds of daily posts. The workflow template library has thousands of community-shared workflows you can import. Discord is busy. Job titles like "n8n developer" and "n8n automation consultant" are appearing on LinkedIn — the platform has crossed the threshold where having it on your resume is a credential.
The release cadence is fast — new nodes ship most weeks. The current focus areas in 2026 are deeper LangChain coverage, MCP (Model Context Protocol) support, multi-tenant Enterprise features, and tighter SaaS company embedding (white-label n8n inside your product).
Frequently asked questions
What is n8n in simple terms?
n8n is a workflow automation platform that lets you connect apps, APIs, and AI models visually — without writing code. Think 'Zapier you can install on your own server' with deeper integrations and AI agent building blocks. The Community Edition is free and source-available; n8n Cloud is paid managed hosting starting at €24/month.
How is n8n pronounced?
n8n is pronounced 'n-eight-n.' It is short for 'nodemation' — a portmanteau of Node.js (the runtime n8n is built on) and 'automation.'
Is n8n free?
Yes. n8n Community Edition is free, source-available under a fair-code license, and runs on your own servers with no usage limits. n8n Cloud (managed hosting) is paid — starting at €24/month — but offers a 14-day free trial without a credit card.
What can you build with n8n?
Almost any workflow that involves moving data between services. Common examples: lead enrichment pipelines, AI customer support agents, RAG (retrieval-augmented generation) over a knowledge base, ETL into a data warehouse, CRM hygiene routines, internal admin tools, multi-channel alerting, and automated reporting.
Is n8n open source?
n8n is source-available under a 'fair-code' license called the Sustainable Use License. You can read the source, self-host, modify it, and use it commercially — but you cannot resell n8n as a hosted service. Strictly speaking it is not OSI-approved open source, but it is functionally equivalent for most users.
Who owns n8n?
n8n is built by n8n GmbH, a venture-backed company headquartered in Berlin, Germany. It was founded in 2019 by Jan Oberhauser. As of 2026, the company has raised a Series B from Highland Europe and supports a team of 200+ engineers.
Sources & further reading
- Official n8n website — n8n.io
- n8n documentation — docs.n8n.io
- n8n on GitHub — github.com/n8n-io/n8n
- Sustainable Use License explainer — docs.n8n.io/sustainable-use-license/
- Zapier's own n8n comparison — zapier.com/blog/n8n-vs-zapier