n8n Review 2026: The Workflow Platform SAP Just Valued at $5.2 Billion
Most people still describe n8n as "the open-source Zapier alternative." That undersells what's happened since: a €55M Series B, a $180M Series C, and in May 2026 a SAP strategic investment that doubled n8n's valuation to $5.2B and embedded it directly inside SAP Joule Studio. Along the way, n8n crossed $100M in annual recurring revenue and became genuinely AI-native. This review covers what Starter/Pro/Business/Enterprise actually cost today, what changed, and whether the fair-code license and steep learning curve are worth it.
| License model | Fair-code (Sustainable Use License) |
| Free self-hosted tier | Community Edition, unlimited executions |
| Cheapest paid plan | Starter, €20/mo billed annually |
| Latest valuation | $5.2B (SAP-backed, May 2026) |
| ARR | $100M (Apr 2026), up from $70M (Dec 2025) |
n8n didn't stay a scrappy self-hosted tool. SAP just bet $5.2B that it's infrastructure.
"The Community Edition — free, self-hosted, unlimited executions — still exists and still works exactly like it always did. What's new is everything built around it: a $100M ARR business, a SAP partnership that embeds n8n inside Joule Studio, and four cloud tiers priced by monthly execution volume rather than by seat. If you tried n8n in 2023 as a self-hosted Zapier substitute and haven't looked since, the honest update is that it's now genuinely AI-native — and genuinely more expensive to run at scale if you're not paying attention to your execution count."
n8n's core interaction model hasn't changed: a node-based canvas where you wire triggers to actions, with full JavaScript/Python code access when the pre-built nodes run out. What changed is the business and the AI layer wrapped around it — LangChain-based AI Agent nodes, an MCP Server Trigger that turns any workflow into a tool an AI client can call, and a valuation that just doubled in a month.
A node-based canvas, a fair-code license, and now an AI orchestration layer.
n8n is a workflow automation platform: you connect triggers (webhooks, schedules, app events) to actions (API calls, data transforms, AI model calls) on a visual canvas, with the option to drop into raw JavaScript or Python whenever a built-in node isn't enough. It ships as a free self-hosted Community Edition or as four metered Cloud tiers, and as of 2026 more than 80% of workflows built on the platform involve an AI agent somewhere in the chain.
The one asterisk worth flagging early: n8n is fair-code, not OSI-approved open source. The Sustainable Use License lets you self-host and modify it freely for internal use, but you cannot resell it as a competing hosted service without a commercial agreement.
n8n's real 2026 story is SAP, not the star count.
n8n-io/n8n, verified on GitHub, Sep 22, 2026.
SAP strategic investment, May 12, 2026.
April 2026, up from $70M end of 2025.
Series B → Series C → SAP, Mar 2025–May 2026.
| Date | Event | Result |
|---|---|---|
| Mar 2025 | Series B, €55M | Valuation reaches €300M |
| Oct 9, 2025 | Series C, $180M, led by Accel | Total funding $240M; valuation $2.5B |
| May 12, 2026 | SAP strategic investment, >€60M | Valuation doubles to $5.2B; n8n embeds natively in SAP Joule Studio (GA targeted Q3 2026) |
The praise is about flexibility. The complaints are about the learning curve and the canvas.
Users consistently praise the flexibility and depth of integrations, calling the visual builder intuitive for both technical and non-technical people — but name a steep learning curve, vague debugging messages, and silent node failures as the recurring first-week pain.
"The canvas becomes hard to follow once a workflow grows beyond roughly 40-50 nodes" — a recurring theme once teams move past simple automations into multi-branch AI agent workflows.
Independent and competitor-published comparisons agree on the shape of the trade-off: "n8n wins on data ownership, depth, and AI-native workflows" while presenting "the most technical interface of the three" automation tools.
Four Cloud tiers, priced by executions — plus a genuinely free self-hosted path.
Community
Unlimited executions, your infra.
- Unlimited workflows & users
- Full node library
- Community forum support
Starter
2.5K executions/mo, 5 concurrent.
- 1 shared project
- 2,300 AI credits/mo
Pro
10K executions/mo, 20 concurrent.
- Admin roles, workflow history
- Up to 13,700 AI credits/mo
Business
40K executions/mo. Self-hosted option available.
- SSO/SAML/LDAP, Git version control
- AI Assistant (preview)
Enterprise: custom pricing, unlimited shared projects, 200+ concurrent executions, external secret store, dedicated SLA. Source: n8n.io/pricing, verified Sep 22, 2026.
Your answer depends on who's running your infrastructure.
Community Edition, self-hosted — genuinely free, unlimited executions.
Starter at €20/mo, billed annually.
Pro at €50/mo covers 10K executions with admin roles and history.
Business at €667/mo (or 50% off if you're a startup under 20 people).
Enterprise, custom pricing.
The thing n8n leaned hardest into in 2026: building and running AI agents, not just gluing APIs together.
n8n's LangChain-based AI Agent node lets you attach tools, memory, and a model (OpenAI, Anthropic, Google, or a self-hosted Ollama model) directly onto the canvas, alongside 70-plus purpose-built AI nodes for RAG, vector stores, and chat triggers. By late 2025, n8n reported that more than 80% of workflows built on the platform involved an AI agent somewhere in the chain — a genuine shift from its original "automate this API call" identity.
- The AI Agent node handles tool-calling, memory, and multi-step reasoning without leaving the canvas.
- You can point any AI node at a self-hosted Ollama model instead of a paid API — a genuine cost lever.
- An in-app AI Workflow Builder can draft a starting workflow from a plain-language prompt.
- Each AI Agent step still consumes one execution and (on Cloud) AI credits — complex agent loops burn through your monthly quota fast.
- Model output quality depends entirely on the model you connect; n8n doesn't improve a weak model's judgment.
A real canvas, and a five-step lifecycle underneath it.
Official n8n video: building your first AI agent from a blank canvas.
Published by the official "n8n" YouTube channel (youtube.com/@n8n-io).
The Community Edition is real, and it scales further than most "free tiers" do.
Self-hosted n8n (Community or Business) runs on infrastructure you control — Docker, Kubernetes, or bare metal — with no execution cap and no per-workflow fee. Production deployments typically move to queue mode: a main n8n instance handles the editor and webhooks while separate worker processes pull executions off a Redis queue, letting you scale workers horizontally as volume grows.
n8n's MCP Server Trigger turns any workflow into a tool an AI client can call.
Since a public-preview launch on April 29, 2026 — available across Cloud, Enterprise, and Community editions alike — the MCP Server Trigger node exposes a single workflow's tools to any external MCP client. Since the same update, it can also build and update workflows, not just trigger them, meaning a client like Claude can create or modify n8n automations through the same protocol. An optional Instructions field was added August 18, 2026 to steer how connected agents use each exposed tool.
You pay per execution, not per step — and there's no default spend cap.
| What you're doing | What you pay |
|---|---|
| Self-hosted Community Edition | €0 to n8n; your own infrastructure cost only |
| Cloud Starter | €20/mo for 2.5K executions; one execution = one full workflow run, regardless of step count |
| Cloud at scale (50K executions/mo) | Independent comparisons put n8n Cloud around $70/mo for this volume, versus $900+/mo on Zapier for the same run count |
| AI credits (Starter/Pro) | 2,300–13,700 AI credits/mo included, consumed by AI-node calls on top of your execution count |
Where your data lives is a deployment choice, not a fixed policy.
People who need logic and code, not just app-to-app triggers.
| Situation | Why n8n fits | Likely plan |
|---|---|---|
| Solo developer automating personal/small-business tasks | Self-host for free, unlimited executions | Community Edition |
| Small team needing a hosted plan without infra ops | No server to manage, fast setup | Starter or Pro |
| Team building AI agents into daily workflows | LangChain AI Agent node, MCP Server Trigger, 70+ AI nodes | Pro or Business |
| Regulated org needing SSO/data residency control | Self-hosted Business/Enterprise, region-pinned | Business or Enterprise |
| Non-technical team wanting the simplest possible UI | n8n is the most technical of the major automation tools | Consider Zapier or Make instead |
Four real gaps, not manufactured ones.
n8n is fair-code (Sustainable Use License), not OSI-approved open source — you can self-host and modify freely for internal use, but not resell it as a competing hosted service.
A recurring, sourced complaint: large AI-agent workflows get visually hard to follow without deliberate use of sub-workflows and node groups.
One execution counts per full workflow run regardless of step count, but there's no built-in ceiling — a bad loop or a stuck AI agent can burn through a monthly quota before you notice.
n8n's own materials cite Microsoft, KPMG, Vodafone, Delivery Hero, Volkswagen, and others among "3,000+ enterprise customers." These are vendor-published claims; JAVIS did not independently verify deployment scale at any of them.
If n8n's technical depth isn't what you need, here's how to think about it.
| If you mostly need... | Compare n8n with... |
|---|---|
| The simplest possible setup, largest app catalog, non-technical team | Zapier (zapier.com) — premium pricing, easiest onboarding |
| A visual builder cheaper than Zapier but simpler than n8n | Make (make.com) — middle ground on price and complexity |
| A free local model to power n8n's AI Agent node without per-token fees | Ollama — read our review |
| A chat interface to sit in front of the models n8n orchestrates | Open WebUI — read our review |
n8n runs a real affiliate program. JAVIS has not enrolled in it.
n8n operates a public affiliate program (30% commission on Cloud referrals for 12 months, via PartnerStack) that any creator can apply to. JAVIS has not joined that program and holds no tracking link for it — every CTA in this article points to n8n's official site, unmodified.
Go to the official page.
Open n8nRead the alternative that matches your real question.
Choose the next articlen8n pairs naturally with the local-AI stack it can orchestrate.
Questions people are actually searching right now
Is n8n free?
Yes, if you self-host: the Community Edition is free with unlimited executions on your own infrastructure. The Cloud plans (Starter, Pro, Business, Enterprise) are hosted and metered by monthly execution count.
Is n8n open source?
Not in the OSI-approved sense. n8n is "fair-code," distributed under the Sustainable Use License — free to self-host, modify, and use internally, but you can't resell it as a competing hosted service without a commercial agreement.
What did SAP just do with n8n?
In May 2026, SAP made a strategic investment of more than €60M in n8n, doubling its valuation to $5.2B, and is embedding n8n natively inside SAP Joule Studio, its AI agent-building environment — general availability was targeted for Q3 2026.
How much does n8n Cloud cost?
Starter is €20/month billed annually for 2.5K executions; Pro is €50/month for 10K executions; Business is €667/month for 40K executions with a self-hosted option; Enterprise is custom. All tiers now include unlimited workflows and users as of April 2026.
Is n8n better than Zapier or Make?
Different trade-offs: Zapier wins on simplicity and the largest app catalog; Make sits in the middle on price and ease; n8n wins on depth, AI-native workflows, and cost at high execution volumes, at the price of a steeper, more technical interface.
Does n8n support MCP?
Yes — the MCP Server Trigger node (public preview since April 29, 2026) turns any workflow into a tool an external MCP client like Claude can call, and since the same update it can build and update workflows, not just trigger them.
Can n8n run local AI models?
Yes. n8n's AI nodes can connect to a self-hosted Ollama instance instead of a paid model API, which removes per-token cost for the AI portion of a workflow (you still pay for the execution itself on Cloud plans).
How is n8n's usage billed?
By execution, not by step: one full workflow run counts as one execution regardless of how many nodes it contains. There is no default spending cap, so runaway loops or stuck AI agents can consume your quota quickly.
Where this came from, and when it was checked.
GitHub stats (stars/forks/license): GitHub data for github.com/n8n-io/n8n, Sep 22, 2026.
Pricing (Community, Starter, Pro, Business, Enterprise): n8n.io/pricing, official, read directly, Sep 22, 2026.
Funding timeline (Series B, Series C, SAP investment): blog.n8n.io/series-c/ and blog.n8n.io/n8n-sap/, official n8n blog, corroborated by PR Newswire/TradingView coverage.
ARR figures ($70M end 2025, $100M April 2026): reported via company-sourced coverage of n8n's growth; not independently audited by JAVIS.
MCP Server Trigger: docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-langchain.mcptrigger and blog.n8n.io, official.
Sustainable Use License: docs.n8n.io/privacy-and-security/sustainable-use-license, official.
Delivery Hero case study: n8n.io/case-studies/delivery-hero/, official, with quote attributed to Dennis Zahrt, Director of Global IT Service Delivery.
Real screenshots/media: n8n.io official OG image, opengraph.githubassets.com repo card, and four screenshots from the official n8n-io/n8n-docs repository (canvas-groups-overview.png, ai-workflow-builder.png, build-n8n-workflow-claude.png, MCP-Server-Trigger.png).
Official video: youtube.com/watch?v=GuaKeDS6UKU, authorship confirmed on YouTube (author "n8n", channel @n8n-io).
User sentiment: G2/Capterra aggregate ratings; independent n8n-vs-Zapier-vs-Make comparisons including Zapier's own published piece; community patterns on canvas scaling and learning curve.
