n8n

n8n is a Berlin-built workflow automation platform used by 1.7M monthly builders and 1,400+ enterprises. It combines a visual node editor with JavaScript/Python code, native AI agent tooling, MCP support, and execution-based cloud pricing. Self-host the Community Edition for free or run managed Cloud plans starting at €20/month.

What sets it apartTechnical, self-hostable AI workflow automation with code-friendly nodes, MCP support, human-in-the-loop controls, and execution-based cloud economics—now embedded inside SAP Joule Studio.

Quick Verdict: n8n in July 2026

n8n is the most technical option in the mainstream workflow automation market. It’s the only major player that gives you a real self-host path, code steps, MCP, multi-agent AI tooling with human-in-the-loop, and execution-based pricing—all from a single Berlin-built vendor that’s now valued at $5.2 billion after a SAP strategic investment in May 2026.

If your team is comfortable debugging APIs and you care about cost predictability, data sovereignty, or building AI agents that actually do things, n8n is the obvious pick. If you want a marketer-friendly tool with zero code, Zapier or Make will feel friendlier.

What n8n Is (and Isn’t) in 2026

n8n is a node-based workflow automation platform built on Node.js and TypeScript (Wikipedia, July 2026). Each node is an action—HTTP request, SaaS connector, AI model, code step, database query, webhook—and you wire them on a visual canvas. A single run of the whole graph counts as one execution.

What it’s not: it’s not a pure no-code tool. Code isn’t optional if you want to go deep, and the DataCamp n8n vs Zapier comparison calls that out plainly. It’s also not a Zapier clone—n8n charges by full execution, not by individual task, and it ships a self-hosted Community Edition under its Sustainable Use License, which the company calls “fair code.”

By July 2026, the n8n homepage reports roughly 198,800 GitHub stars, 4.7/5 on G2, and over 200,000 community members. A May 2026 Highland Europe press release pegs the active builder base at 1.7 million monthly users and 1,400+ enterprise customers.

“n8n was the big unlock. Tools like ChatGPT and Claude are great, but n8n is the thing that allows you to integrate AI into your work and your processes in a safe and controlled way.” — Ollie Scheers, CTO at Huel (n8n case study)

Who Should Use n8n in 2026

Pick n8n if any of these sound like you:

1. Developers and platform engineers who want a visual builder that still lets them drop into JavaScript or Python code at any node (n8n Features page).
2. AI builders who need MCP, multi-agent orchestration, RAG templates, and human-in-the-loop gates before their agent sends an email or updates a CRM (blog.n8n.io Production AI Playbook).
3. Ops, security, and IT teams who need to keep data inside their own perimeter, integrate with HashiCorp Vault or AWS Secrets Manager, and run workflows on infrastructure they control (n8n Enterprise, n8n docs on external secrets).
4. Startups whose per-task costs on Zapier or Make are ballooning as workflows get more complex (blog.n8n.io execution advantage).
5. Anyone fed up with monthly task surprises—the n8n docs make it clear that a 2-step workflow and a 200-step AI agent both count as a single execution.

Skip n8n if you’re a non-technical marketer who wants the simplest connector on earth with no code and zero interest in self-hosting. You’re going to be happier in Zapier or Make.

Latest Features and 2026 Updates

n8n ships a new minor release most weeks. The docs changelog puts the current stable at 2.32.6 and beta at 2.33.2 as of late July 2026. Here’s what actually moved in 2026:

• July 2026 (v2.30): App-only authentication for Microsoft Excel 365, Teams, and To Do via Microsoft Entra Service Principal, plus mTLS support for Kafka triggers and credentials.
• June 2026 (v2.28–v2.29): Canvas Groups to collapse related nodes into named blocks (Ctrl/Cmd+G); MCP server gains custom/community node support, editable workflow settings, and version-history rollback; GitHub App authentication replaces PATs; built-in web search for AI agents that falls back to Brave or a self-hosted SearXNG.
• May 2026 (v2.20–v2.21): The Microsoft Agent 365 Trigger node lets you publish an n8n agent as a member of your Microsoft 365 tenant with its own email and Entra identity. Verified webhook signatures across 14 trigger nodes (Acuity, Asana, Calendly, Twilio, etc.).
• April 2026 (v2.17–v2.19): Moonshot Kimi and Alibaba Cloud Model Studio join as native AI providers; IdP role mapping now configurable inside n8n; instance bootstrapping via env vars for SSO and security policies.
• March 2026 (v2.13–v2.16): Visual diff in workflow version history, Oracle DB vector store, 1Password as an external secrets provider, easier credential setup on Cloud via managed OAuth for Slack, GitHub, HubSpot, etc.
• February 2026 (v2.8): Personal space policies for instance admins—publishing and sharing controls for end users.
• January 2026 (v2.5–v2.6): Human-in-the-loop gates on AI tool calls and a new Chat node with “send and wait for response” operations. This is the release that makes agentic workflows actually safe to put in production.

The May 2026 SAP partnership is the biggest non-product news: n8n will live natively inside SAP’s Joule Studio agent builder inside SAP Business AI Platform.

Pricing Breakdown July 2026

Cloud pricing is listed in euros on n8n.io/pricing (USD-equivalents fluctuate with FX). The Community Edition is free to self-host—you only pay infrastructure.

Plan	Price (billed annually)	Executions / month	Concurrent runs	Best for	AI Assistant credits
Community (self-hosted)	Free + infra	Unlimited (your hardware)	Your hardware	Hobbyists, evaluators, learning	N/A
Starter (Cloud)	€20 / mo	2,500	5	Solo builders, pilots	2,300 / month
Pro (Cloud)	€50 / mo	10,000	20	Production teams	up to 13,700 / month
Business (self-hosted w/ license)	€667 / mo	40,000	200	Mid-market, <100 employees	Coming soon
Enterprise	Custom	Custom	200+	Regulated industries, Fortune 500	Coming soon

A few pricing details worth knowing, all from the n8n pricing FAQ:

• All plans include unlimited users and unlimited workflows. Only executions are metered.
• AI Assistant is still labeled preview. You can’t yet buy extra credits—if you run out, you upgrade the plan.
• Overage on Business: extra buckets of 300,000 executions cost €4,000.
• Free trial: 14 days on Starter, Pro, and Business—no credit card for Starter and Pro.
• Wire transfer and invoice billing are available on the annual Business plan and Enterprise (per the pricing FAQ).

A useful mental model: per the n8n blog pricing explainer, a workflow with 100,000 internal tasks could cost over $500/month on per-task tools; on Pro it’s still one execution per run.

n8n vs Alternatives (Comparison Table)

Here’s how n8n stacks up against the most common alternatives a technical buyer would consider. Pricing column reflects entry-level paid cloud tiers as of July 2026.

Tool	Pricing model	Self-host?	Pre-built integrations	Code-first?	AI agent tooling	Best fit
n8n	Per full execution	Yes (Community free)	1,000+	Yes (JS/Python)	Strong (Agents, MCP, HITL)	Technical teams, AI builders
Zapier	Per task	No	8,000+	Limited (Code by Zapier)	Basic AI actions	Non-technical teams, broad SaaS
Make	Per operation/credit	Cloud only (on-prem Agent for enterprise)	2,800+	Custom JS on Enterprise	AI modules	Visual power users, mid-market
Workato	Per recipe/task	No (cloud)	1,000+	Limited	Recipe-based AI	Enterprise IT integration
Pipedream	Per credit / invocation	Limited	2,500+	Yes (Node.js/Python)	Workflow agents	Developer-first serverless style

What jumps out: only n8n and Pipedream really embrace code. Only n8n gives you a free self-hosted edition with governance features unlocked by a license key. And only n8n pairs AI agents with built-in human-in-the-loop gating at the tool-call level.

For a deeper head-to-head, the official n8n vs Zapier and n8n vs Make pages lay out the arguments from n8n’s side, and the DataCamp n8n vs Zapier piece gives a more neutral walkthrough.

Real-World Use Cases (and What Customers Actually Got)

Numbers below come from n8n’s public case studies and the Highland Europe Series B press release.

1. Huel (nutrition brand, ~350 employees) rolled n8n out enterprise-wide in nine months. CTO Ollie Scheers’ team built ~200 live workflows, saved about 1,000 hours of manual work, and cancelled roughly £100,000 of annual SaaS licenses by replacing them with n8n + Airtable. (Huel case study)
2. Vodafone UK replaced a Tines and IBM Resilient pilot with n8n for SOAR. Since August 2024 they launched 33 workflows, saved 5,000+ person-days, and avoided £2.2M in costs—about £300K/month in ongoing savings. (Vodafone case study)
3. Fullscript has 200+ employees (30%+ of the org) building on n8n, with 23 workflows in production. Their security-investigation workflow reduced investigation time by 97% and is on track to save 3,600 engineering hours a year. (Fullscript case study)
4. Delivery Hero reportedly saves 200 hours/month and Musixmatch saved 47 days of engineering work in four months, per the Highland Series B release.

Those are concrete workflows—customer-support email triage, Slack-based security bots, SOAR enrichment pipelines, RAG over internal PDFs—built by non-engineers once the patterns are set.

Limitations and Gotchas I Wish Someone Told Me Earlier

n8n is powerful, but it’s not magic. Here’s where teams actually trip up:

• Steep curve for non-developers. The TrustRadius review by Adrian Tsallis (TrustRadius, July 2025) is typical: “the learning curve is a bit too steep for beginners when tackling complex problems.”
• Self-hosting means real ops. You own upgrades, backups, TLS, database scaling, and security hardening. Per the n8n security docs, encryption at rest on self-hosted is your job.
• Cloud prices are in EUR. Plan budgets assuming currency conversion and FX noise.
• AI Assistant is preview and credit-capped. You can’t yet top up credits mid-month on cloud—you upgrade. Self-hosted AI Assistant, which lets you bring your own model API key, is “coming soon” per the pricing FAQ.
• Some “Enterprise” features are Enterprise-only. SSO/SAML, IdP role mapping, audit log streaming to Datadog/Splunk, and redaction of execution data still require the top tier.
• Community Edition license is fair-code, not OSI open source. You can self-host and modify freely, but you can’t resell a hosted n8n product commercially without a separate agreement (see the Sustainable Use License).
• Cloud region is currently EU-only. Per the pricing FAQ, additional regions are being prepared.

Privacy, Security, and Compliance

This is one area where n8n punches above its weight.

• SOC 2 Type 2 audited, with annual external audits and quarterly vulnerability scans. The SOC 3 report is publicly downloadable from the n8n Security page.
• GDPR-aligned, with a dedicated EU Data Protection Officer and detailed lawful-basis tables in the Privacy Policy (last updated 24 June 2026).
• Cloud hosting on Microsoft Azure Germany West Central (Frankfurt), with TLS via Cloudflare, AES-256 encryption at rest, private networking, daily encrypted backups, and at least 12 months of audit log retention.
• External secret stores for HashiCorp Vault, AWS Secrets Manager, Azure Key Vault, GCP Secret Manager, and 1Password Connect—see the external secrets docs.
• SSO, SAML, and LDAP on Business and Enterprise plans.
• Execution data redaction for sensitive workflows on Enterprise—manual runs stay visible to developers while production data is hidden by default.
• Human-in-the-loop on AI tool calls, added in v2.6 (Jan 2026), so an AI agent cannot fire off a destructive action without an explicit human approve.
• No training on your data. The Privacy Policy is explicit: “We do not use personal data processed in connection with the Website or Services, including data from third-party services or integrations, to train n8n or third party machine learning (ML) models.”

If you have to defend an automation stack to a CISO, n8n’s Trust Center plus the docs above should get you most of the way.

FAQ

What is n8n and how does it work?

n8n is a node-based workflow automation platform where you visually connect apps, APIs, and AI models on a canvas. Each node is an action—HTTP call, SaaS connector, code, AI model, database—and the whole graph runs as a single execution. You can self-host the Community Edition or use n8n Cloud.

How much does n8n cost in 2026?

n8n Cloud starts at €20/month for Starter (2,500 executions), €50/month for Pro (10,000 executions), €667/month for Business self-hosted (40,000 executions), and custom pricing for Enterprise. The Community Edition is free; you only pay infrastructure. All paid plans include unlimited users and unlimited workflows. Source: n8n pricing.

Is n8n better than Zapier for developers?

For technical teams, generally yes. n8n gives you code steps, self-hosting, MCP, AI agents with human-in-the-loop gating, and execution-based pricing. Zapier wins on the size of its connector library (8,000+) and ease for non-developers, but charges per task and is cloud-only. See the DataCamp comparison for an independent take.

Can I self-host n8n for free?

Yes. The Community Edition on GitHub is free under the Sustainable Use License. You bring the server, database, and ops. To unlock SSO, Git-based environments, RBAC, queue mode, and external secret stores on a self-hosted instance, you add a Business or Enterprise license key.

Does n8n work with AI agents and MCP?

Yes. n8n ships AI Agent nodes, LangChain-style tool orchestration, an MCP Client node and an MCP Trigger for turning any workflow into an MCP server, RAG templates, and Human-in-the-Loop approval gates on tool calls. The AI Assistant for building is still in preview on cloud and uses credits; self-hosted AI Assistant is coming soon.

Where is n8n Cloud data stored?

n8n Cloud runs on Microsoft Azure in the Germany West Central region (Frankfurt, EU). Data is encrypted at rest with AES-256, in transit with TLS via Cloudflare, and backed up daily to a separate region in the same country. Source: n8n Security.

Is n8n open source?

It’s fair-code, not OSI open source. The source is public on GitHub, you can self-host and modify it freely, and contributions are welcomed. Reselling n8n as a hosted commercial product requires a separate license. The license is documented at docs.n8n.io.

Who owns n8n and is it well-funded?

n8n GmbH is a private Berlin company founded by Jan Oberhauser in 2019. Funding milestones: $1.5M seed in March 2020 (Sequoia, TechCrunch), $12M Series A in April 2021 (Felicis Ventures, TechCrunch), €55M Series B in March 2025 (Highland Europe, press release), $180M Series C in October 2025 at a $2.5B valuation (Accel-led, Wikipedia), and a SAP strategic investment in May 2026 that doubled its valuation to $5.2B (Highland Europe).

How does n8n handle scaling?

n8n supports queue mode (main + workers), multi-main for high availability, and horizontal scaling on Docker Compose or Kubernetes. The docs cite benchmark numbers around 220 workflow executions per second on a single instance. See scaling docs and queue mode docs.

What integrations does n8n support?

The n8n integrations directory lists 1,000+ nodes covering SaaS apps, databases, AI providers, and community-built connectors, plus generic HTTP, GraphQL, and webhook nodes that can reach any API. Verified partner nodes include SerpApi, ElevenLabs, Brave Search, Qdrant, and Bright Data. MCP lets you expose any n8n workflow as a tool to Claude, Lovable, or other MCP clients.

Final Verdict

n8n in mid-2026 is the rare automation tool that gets more useful as your workflows get more complex, not less. The execution-based pricing means an ambitious AI agent pipeline costs the same as a two-step webhook. Self-hosting means you can build something that survives a vendor outage, an audit, or a regional data residency requirement. And the AI tooling—multi-agent orchestration, MCP, human-in-the-loop gating—actually holds up in production because n8n treats AI like a workflow step, not a black box.

The trade-offs are real: it’s not the friendliest tool for non-technical builders, self-hosting is real work, and the AI Assistant is still preview. If that’s not your problem to solve, you can ignore most of those caveats.

Buy n8n if you want full control, real AI orchestration, and pricing that scales with outcomes rather than steps. Look at Zapier or Make if you want pure ease of use and a non-technical builder is your primary user. Look at Pipedream if you want a developer-first serverless feel but don’t need self-hosting. Look at Workato if you’re an enterprise IT team standardizing on a managed iPaaS.