n8n Review 2026: Pricing, 1,700+ Nodes & Self-Hosting Tested

What is n8n?

n8n (pronounced "n-eight-n", short for "nodemation") is a source-available workflow-automation platform. You build automations visually by connecting nodes on a canvas — a trigger node starts the workflow, and each subsequent node fetches data, transforms it, calls an API, or runs custom code. Where it separates itself from mainstream tools like Zapier and Make is depth: any node's output is fully inspectable, you can drop into JavaScript or Python at any step, and you can run the entire engine on your own infrastructure for free.

That combination — visual for speed, code for anything the visual layer can't express, self-hostable for control and cost — is why n8n has become the automation backbone for a large share of technical teams and the fast-growing crowd building production AI agents. It is licensed under the fair-code Sustainable Use License: the source is public, you may self-host and modify it freely, and the only real restriction is that you can't repackage n8n and resell it as a competing hosted service.

Key features

1,700+ integrations and a universal HTTP node

n8n ships with more than 1,700 pre-built integration nodes covering the usual suspects — Google Workspace, Slack, Notion, Airtable, HubSpot, Postgres, AWS — plus hundreds of niche tools. Crucially, when a dedicated node doesn't exist, the generic HTTP Request node can call any REST or GraphQL API, so you are never blocked by a missing connector the way you can be on closed platforms. This is the single biggest practical advantage for teams integrating internal or long-tail services.

70+ native AI and LangChain nodes

n8n has invested heavily in AI. It includes 70+ native LangChain nodes covering chat models (OpenAI, Anthropic, Google, local models via Ollama), embeddings, vector stores (Pinecone, Qdrant, Supabase, PGVector), memory, output parsers, and tool-use routing. The dedicated AI Agent node lets you build an agent that reasons over a goal and calls other n8n workflows as tools — turning the platform into a practical orchestration layer for retrieval-augmented generation and multi-step agents, not just simple "if this then that" automation.

Code when you need it

The Code node runs JavaScript or Python against the data flowing through a workflow, so transformations that would require awkward multi-step chains on other platforms become a few lines. You can also install npm packages on a self-hosted instance. This escape hatch is what lets one n8n workflow replace what might be three or four brittle Zaps.

Self-hosting and deployment

The Community Edition runs on Docker, Kubernetes, or a plain Linux VM in minutes. Self-hosting means your data never leaves your infrastructure — a decisive factor for teams with compliance or data-residency requirements — and there are no per-execution fees, so high-volume automations cost only the compute you already pay for. Enterprise self-hosted adds SSO, RBAC, audit logs, log streaming, and external secret stores.

n8n pricing in 2026

n8n's pricing has two completely different tracks: free-forever self-hosting, or managed cloud billed by workflow executions (a full run of a workflow), not by individual steps. That per-execution model is a big deal — a 20-step workflow and a 2-step workflow both count as one execution, so complex automations don't get taxed the way step-based competitors charge. Figures below are the annual-billing rates verified against n8n's official pricing page; month-to-month billing runs roughly 20% higher.

Every tier — including free — includes unlimited users and workflows and every integration; you're only ever paying for execution volume and enterprise governance features. There's also a Startup plan offering 50% off Business for companies under 20 employees. For a fuller cost breakdown across automation tools, see our AI pricing guide and the deep-dive on n8n pricing for 2026.

n8n vs Zapier vs Make

Zapier wins on breadth of app catalog and sheer ease — if a non-technical marketer needs to connect two SaaS tools this afternoon, Zapier is the fastest path, but its per-task pricing gets expensive quickly and its logic ceiling is low. Make (formerly Integromat) sits in the middle with a more visual, branching canvas and better value than Zapier, but it is still fully hosted and closed. n8n is the technical team's choice: more powerful, dramatically cheaper at volume (especially self-hosted), and the only one of the three you can run entirely on your own servers. For the detailed head-to-heads, read n8n vs Zapier, n8n vs Make, the three-way n8n vs Make vs Zapier, and n8n vs Node-RED.

Who should use n8n?

Choose n8n if you have at least one technically comfortable person, you care about data control or self-hosting, you're building AI agents or RAG pipelines, or you're scaling past ~10,000 monthly executions and want to escape runaway per-task pricing. Look elsewhere if your team is entirely non-technical and you value the absolute fastest, most template-driven setup over flexibility and cost — in that case Zapier remains the gentler on-ramp. Most teams that start on Zapier and hit either a cost wall or a logic wall eventually migrate to n8n.

Setup and learning curve

Cloud signup takes minutes; self-hosting via Docker is a single command and roughly 15–30 minutes to a production-ready instance with a reverse proxy. The core concept — trigger node plus a chain of action nodes — is intuitive, but expression syntax, error-handling branches, and the Code node take a few days to feel fluent. For engineers building AI workflows specifically, our n8n API guide covers the patterns worth learning first. Budget a weekend to get genuinely productive; the payoff is automations you simply cannot build on step-based competitors.

Final verdict

n8n earns a 9.0/10 as the most capable and best-value automation platform for technical teams in 2026. The free self-hosted edition alone would justify a look; the depth of AI tooling and the execution-based pricing make it the strategic long-term choice for any team serious about automation and AI agents. Its only real weakness is the learning curve relative to consumer-grade tools — a fair price for the ceiling it removes. Scores reflect independent editorial testing; see our review methodology.

Frequently asked questions