n8n

TL;DR

• What it is: A fair-code workflow automation platform with 400+ integrations, native AI/LangChain support, and full self-hosting capability. Think Zapier, but you own the server and can write JavaScript anywhere in the pipeline.
• Who it’s for: Technical teams that want visual workflow building with escape hatches into code. DevOps, SecOps, IT automation, and AI agent builders who need more than drag-and-drop.
• Cost savings: Zapier Professional runs $49–300+/mo depending on task volume. n8n self-hosted runs on a $6–20/mo VPS with execution-based pricing only on cloud. Self-hosted is free forever under the fair-code license.
• Key strength: The deepest feature set in open-source automation. JavaScript/Python code nodes, LangChain integration, 900+ workflow templates, SSO/RBAC/LDAP for enterprise, and a 200K+ community.
• Key weakness: The “fair-code” license is not truly open-source — you cannot resell or embed n8n commercially without a separate agreement. Hidden operational costs for self-hosting can reach $200–500/mo when you factor in maintenance, security, and monitoring. The integration count (~400) still trails Zapier’s thousands.

---

What is n8n

n8n (pronounced “n-eight-n,” short for “nodemation”) is a workflow automation platform where you connect triggers to actions across hundreds of apps using a visual drag-and-drop editor. New form submission comes in, data flows through transformation nodes, lands in your CRM, fires a Slack notification — that kind of thing. The project was founded in Berlin in 2019 by Jan Oberhauser and has grown to 180K+ GitHub stars, making it one of the top 50 most-starred projects on all of GitHub.

What separates n8n from the Zapier/Make crowd is the code-first escape hatch. Every workflow can contain JavaScript or Python nodes where you write arbitrary logic, import npm packages, and manipulate data however you want. On the self-hosted version, you get full npm package access — something the cloud version restricts. This makes n8n feel less like a no-code tool and more like a visual scripting environment that happens to have a clean UI.

The AI angle is real, not marketing vapor. n8n ships native LangChain integration for building AI agent workflows with vector stores, memory, and tool-calling. Vodafone used it to build threat intelligence automation that saved 2.2 million pounds. Huel credits it with saving 1,000 hours of manual work. These are not hypothetical use cases.

The license deserves attention: n8n uses a “Sustainable Use License” that the company calls “fair-code.” It is source-available and self-hostable, but it is not MIT or Apache. You cannot redistribute n8n commercially or embed it in your product without a separate enterprise license. For most self-hosters this doesn’t matter. For anyone building a product on top of it, read the license carefully.

---

Why people choose it over Zapier, Make, and Activepieces

Versus Zapier

The comparison everyone makes, and the one where n8n wins on substance. Zapier charges per task — the more your workflows fire, the more you pay. At scale, this becomes punitive. Zapier Professional starts at $49/mo for 2,000 tasks and climbs rapidly into hundreds per month.

n8n self-hosted has zero per-execution cost. You pay for the VPS and your time. The Zapier blog itself acknowledges the distinction, framing it as “n8n for technical teams that want control” versus “Zapier for organizations that want managed AI orchestration.” That framing is generous to Zapier — the real translation is: Zapier is easier to start with and more expensive to scale with.

The integration gap is real, though. Zapier has 3,000+ integrations. n8n has 400+. For mainstream SaaS (Slack, Gmail, HubSpot, Notion), n8n covers you. For niche tools, you’ll fall back to the HTTP request node and webhooks.

Versus Make (Integromat)

Make is the middle ground — cheaper than Zapier, more visual scenario building, still closed-source SaaS with operation-based pricing. Make’s visual debugger with step-by-step bundle inspection is genuinely better than n8n’s execution log for complex scenario debugging. But Make can’t run on your server, and you can’t write arbitrary code inside it.

Versus Activepieces

This is the interesting one. Activepieces is MIT-licensed (truly open source), has a cleaner UI, and charges per active flow rather than per task. Multiple reviewers say the Activepieces UI is more approachable for non-technical users. But n8n has significantly more integrations, deeper AI agent capabilities (LangChain, vector stores, memory management), and a larger community.

The practical translation: Activepieces is what you hand to marketing. n8n is what the engineering team builds production automation on.

---

Features: what it actually does

Core workflow engine:

• Visual drag-and-drop builder with 400+ integrations
• JavaScript and Python code nodes with full npm/pip package access (self-hosted)
• Triggers: webhooks, cron schedules, app events, manual, form submissions
• Branching, loops, error handling, auto-retry, sub-workflows
• 900+ community-contributed workflow templates
• Workflow versioning and rollback
• Execution logging with step-by-step input/output inspection

AI and agents:

• Native LangChain integration for AI agent workflows
• Vector store connections for RAG (Retrieval-Augmented Generation)
• AI memory management across workflow executions
• Human-in-the-loop nodes for approval workflows
• AI evaluation nodes for testing agent accuracy
• Supports OpenAI, Anthropic, and custom model endpoints

Enterprise features:

• SSO (SAML), LDAP, RBAC with granular permissions
• Audit logs and log streaming to SIEM
• Encrypted secret stores
• Git-based version control with environment isolation
• Air-gapped deployment support
• Workflow diffs for code review

Deployment:

• Docker one-liner: docker run -it --rm --name n8n -p 5678:5678 docker.n8n.io/n8nio/n8n
• Also: npx n8n for instant local testing
• Kubernetes via Helm charts
• Cloud offering at app.n8n.cloud

---

Pricing: SaaS vs self-hosted math

n8n Cloud:

• Community (free): unlimited workflows, 5 active, limited executions
• Starter: 20 EUR/mo for 2,500 executions
• Pro: 50 EUR/mo for 10,000 executions
• Enterprise: custom pricing with SSO, audit logs, dedicated support

n8n Self-hosted:

• Software: $0 (fair-code license)
• VPS: $6–20/mo on Hetzner, Contabo, or DigitalOcean
• Your time for maintenance and updates

Concrete savings for a typical setup:

Say you’re running 30 workflows that collectively fire 50,000 times per month. On Zapier, you’re deep into custom enterprise pricing — easily $300–500+/mo. On n8n cloud Pro, 50,000 executions is around 250 EUR/mo. Self-hosted on a $15 Hetzner VPS with 4GB RAM, it’s $15/mo flat. Over a year: Zapier ~$4,800. n8n cloud ~$3,000. Self-hosted ~$180 + your ops time.

The hidden cost reality check:

Production-ready self-hosting requires more than a VPS. Realistic minimum for reliable operation: 4GB RAM with 2 CPU cores, SSL certificates, backup strategy, monitoring, security patching, and PostgreSQL tuning. Total operational costs can reach $200–500/mo for a team that needs reliability. That estimate is high for a solo founder running 20 workflows on a $10 VPS. It’s realistic for a 50-person company running hundreds of workflows where downtime costs money.

---

Deployment reality check

Developer Anthony Sidashin describes deploying n8n on a Hetzner server: roughly 30 seconds for the Docker pull, then 20 minutes configuring nginx reverse proxy with Cloudflare HTTPS. That matches most self-hosting reports — this is not a weekend project, it’s a lunch break project if you know what you’re doing.

What you actually need:

• A Linux VPS with 2–4GB RAM (4GB recommended for AI workflows)
• Docker and docker-compose
• A domain with reverse proxy (Caddy or nginx) for HTTPS
• PostgreSQL (bundled or external)
• Optional: Redis for scaling, SMTP for email triggers

What can go sideways:

• The “fair-code” license means you cannot redistribute n8n or include it in a commercial product. This catches some people off guard.
• External npm packages work on self-hosted but not on cloud, which means workflows you build locally may break on cloud migration.
• The integration count (400+) means you’ll hit gaps. Expect to use the HTTP Request node and webhooks for anything outside mainstream SaaS.
• Updates require manual Docker pulls on self-hosted. No auto-update mechanism.
• The AI features (LangChain, vector stores) require separate infrastructure for the models themselves — n8n orchestrates but doesn’t serve models.

Realistic setup time: 30–60 minutes for a developer comfortable with Docker. Half a day for someone learning as they go.

---

Who should use this (and who shouldn’t)

Use n8n if:

• You’re a technical team building production automation that needs code escape hatches alongside visual building.
• You want AI agent workflows with LangChain, vector stores, and memory management.
• You’re spending $200+/mo on Zapier and want that bill to disappear.
• You’re comfortable deploying and maintaining Docker containers.
• Data sovereignty matters — your workflows and data stay on your infrastructure.

Skip it (pick Activepieces instead) if:

• You need a strictly MIT-licensed automation engine for embedding in your product.
• Your team is non-technical and needs the simplest possible UI.
• You value unlimited executions pricing over per-execution pricing.

Skip it (stay on Zapier) if:

• You have fewer than 10 workflows and Zapier’s pricing isn’t painful yet.
• You need 3,000+ native integrations without configuring HTTP requests.
• Your team won’t touch a command line and you don’t have a technical person to maintain infrastructure.

---

Alternatives worth considering

• Activepieces — MIT-licensed, cleaner UI, unlimited executions model, fewer integrations, less AI depth. The “Zapier replacement for non-technical teams.”
• Zapier — The incumbent. Easiest onboarding, most integrations, most expensive at scale, fully closed source.
• Make (ex Integromat) — Cheaper than Zapier, excellent visual debugging, still closed-source and usage-based.
• Pipedream — Developer-first, code-centric, partially open source. Good for teams that think in code rather than flowcharts.
• Windmill — Script-first automation for engineering teams. Less visual, more infrastructure-as-code.

For the typical self-hoster choosing between n8n and Activepieces: pick n8n if your team writes code and needs depth. Pick Activepieces if your team needs simplicity and a truly open-source license.

---

Bottom line

n8n is the power tool of open-source workflow automation. It does more than any competitor in the self-hosted space — AI agents, LangChain, code nodes, enterprise governance — and it has the community and case studies to prove it works at scale. The trade-offs are real: the license isn’t truly open-source, self-hosting isn’t truly free when you count operational costs, and the integration catalog still trails Zapier’s. But for technical teams spending real money on Zapier or Make and tired of per-task pricing anxiety, the math is clear. A $15/mo VPS replaces a $300/mo SaaS bill, and you get to write JavaScript wherever you want.

If the deployment is the blocker, that’s exactly the kind of work that upready.dev handles. One-time setup, you own the infrastructure.

Sources

This review synthesizes 6 independent third-party articles along with primary sources from the project itself. Inline references throughout the review map to the numbered list below.

1. [1] pixeljets.com by Anthony Sidashin (2024-05-19) — “My experience using n8n, from a developer perspective” — tutorial (link)
2. [2] infralovers.com by Paul Strebenitzer (2025-05-09) — “n8n Guide: Self-Hosted Workflow Automation vs Zapier & Make (2025)” — comparison (link)
3. [3] 0hands.com by Unknown (2025-01-01) — “n8n Review 2025: What Can This Open-Source iPaaS Tool Do?” — overview (link)
4. [4] hostinger.com by Unknown (2025-02-18) — “What is n8n? Understanding the workflow automation tool” — tutorial (link)
5. [5] latenode.com by Unknown (2026-02-12) — “N8N Free Self-Hosted Version 2025: Complete Analysis + True Cost Reality Check” — critical (link)
6. [6] zapier.com by Unknown — “Zapier vs. n8n: Which is best for your organization? [2026]” — comparison (link)
7. [7] GitHub repository — official source code, README, releases, and issue tracker (https://github.com/n8n-io/n8n)
8. [8] Official website — n8n project homepage and docs (https://n8n.io)

References [1]–[8] above were used to cross-check claims about features, pricing, deployment, and limitations in this review.