How n8n Turned an Open-Source Side Project Into a $2.5B AI-Automation Machine

Jan Oberhauser spent six years as a solo founder in Berlin building free workflow software almost nobody paid for. Then AI agents happened, revenue went up 10x in a year, and Accel handed him $180M at a $2.5B valuation. The wild part isn't the AI timing. It's that n8n built the entire distribution engine before the money showed up — by giving the product away and letting a community sell it for them.

This is a teardown of how that actually worked. No fluff. Just the moves.

By the numbers

• ▮$240M raised total across five rounds: $1.5M seed (Mar 2020, Sequoia + firstminute), $12M Series A (2021, Felicis), ~$60M Series B (Mar 2025, Highland Europe), $180M Series C (Oct 9, 2025, led by Accel) at a $2.5B valuation.
• ▮10x revenue growth year-over-year and 6x user growth heading into the Series C. ARR had grown past $40M by that round; later reports put it at $100M+ in 2026.
• ▮~196,000 GitHub stars — and they added 112,000 of them in 2025 alone, the #1 JavaScript project on the year's Rising Stars list by a wide margin. (They added ~17k in all of 2024.)
• ▮100M+ Docker pulls. Self-hosting is the real top of funnel.
• ▮400+ integrations and ~10,500 community-built workflow templates on n8n.io.
• ▮80%+ of workflows built on the platform now involve AI agents.
• ▮Near-zero paid search. DataForSEO shows n8n.io pulling ~121k organic visits/month in the US across ~11,000 ranking keywords — and bidding on just 3 paid keywords. They earn traffic, they don't rent it.

The wedge: give away the thing people were forced to rent

Oberhauser had tried the existing automation tools — Zapier, the enterprise iPaaS suites — and found them "either too pricey, or not flexible or robust enough." Worse, they were black boxes. Your data and your logic lived on someone else's server. His line to TechCrunch in 2020: "If the company goes bust or changes policy, you are in trouble."

So the wedge was ownership. n8n shipped free, open, and self-hostable from day one. A developer could run the whole thing on their own box, see every line of code, and never send a byte to a vendor. For technical teams at banks, hospitals, and governments who legally could not use cloud-only tools, that wasn't a nice-to-have. It was the only door in.

Here's the move: don't compete on features against an incumbent. Compete on the thing they structurally can't offer. Zapier couldn't give you the code and the self-host option without cannibalizing its own model. n8n could — because it had no model to protect yet.

Engine 1: the community flywheel (and the license that fed it)

Open-source distribution only works if you don't get strip-mined by a cloud giant reselling your work. n8n's answer was "fair-code" — its Sustainable Use License. The code is visible and free to use, even inside a Fortune 500. But if you want to commercialize n8n as a hosted service, you need a partnership. Open enough to build a community. Closed enough to build a business.

Timing mattered. MongoDB, HashiCorp, and Redis had all burned developers with surprise license switches. n8n showed up with clear, upfront terms and earned trust that was going cheap.

Then they fed the flywheel deliberately. They hired a dedicated community person before there was revenue. They built documentation on a public forum, not a private Slack — so every answered question became a permanent, Google-indexed asset instead of vanishing into a channel. Sixteen thousand developers by the Series A. Over 100 million Docker pulls today.

Steal the mechanic: every support answer should be a public, searchable, reusable page. A forum post compounds. A Slack DM evaporates.

Engine 2: the template library as an SEO printing press

This is the quiet genius. n8n hosts ~10,500 community-contributed workflow templates — "Sync Shopify orders to Google Sheets," "Auto-summarize emails with GPT," one page per use case.

Work the logic:

1. ▮Every template is a landing page targeting a real, high-intent search ("how to connect X to Y").
2. ▮The community writes the pages for free by publishing workflows they already built.
3. ▮Every page is a product demo — the reader is already picturing the automation running.

That's how you get to ~11,000 ranking keywords and ~120k US organic visits a month while bidding on basically no paid terms. The content isn't a cost center; it's user-generated and it converts because the searcher's problem and the product's answer are the same object.

Steal this: if your product does N things, that's N landing pages you don't have to write — if you make it trivial for users to publish what they built.

Engine 3: integrations as distribution, not just features

400+ integrations isn't a feature checklist. Each connector is a new set of keywords, a new template category, a new reason a specific team lands on the site. "n8n Notion," "n8n Slack," "n8n HubSpot" — every integration is another front door indexed by Google, and community members build most of the nodes.

The strategic bet underneath it: stay neutral. n8n never picked a winning app, database, or AI model to bless. It positioned as the connective layer between all of them. When a market is fragmenting fast — and the AI stack in 2023–2025 was fragmenting weekly — the neutral connector wins more shelf space than any single opinionated tool. You don't have to bet on the winner if you plug into all of them.

Engine 4: the AI-agent moment they actually saw coming

Most of n8n's growth is one pivot. Around 2023, Oberhauser watched Pinecone reposition from "vector database" to "the database for AI" and clocked what was about to happen to every infrastructure category. So the team rebuilt n8n's positioning around AI orchestration in roughly six weeks — from "workflow automation" to "the platform where you chain LLMs, tools, and humans into production agents without writing Python."

Revenue 4x'd in the eight months after. Today 80%+ of workflows built on n8n involve AI agents.

Why it landed: n8n already was an agent framework — visual nodes, custom code, 400 tools to call — before anyone used that word. The AI wave didn't require a new product. It required a new sentence. They had spent six years building the substrate; the pivot was recognizing what it had become and renaming it fast.

Steal this: when the market's language shifts, the company that re-describes its existing product first eats the demand. The rebuild was positioning, not engineering. Speed of narrative beat depth of feature.

Engine 5: self-host as the funnel, cloud as the register

The free self-hosted version isn't charity — it's the top of a very long funnel. A developer spins up n8n on a home server for a side project. It works. They bring it to their team. Usage grows. Now they don't want to babysit infra, manage scaling, or handle SSO and audit logs. So they convert to n8n Cloud or an enterprise license.

The product spans that whole range on purpose. n8n deliberately kept a steep learning curve — an embedded JavaScript/Python editor instead of a dumbed-down menu builder — so the same platform serves a solo hacker and a 500-person enterprise without anyone hitting a ceiling and churning to a "real" tool. Power over simplicity. The hobbyist and the buyer use the same product; only the deployment (and the invoice) changes.

Revenue reportedly splits roughly cloud subscriptions ~55%, enterprise licenses ~30%, embedded partnerships ~15% — all fed by that free self-hosted base of 100M+ pulls. In May 2026, SAP took a strategic stake reportedly valuing n8n around $5.2B and embedding the product, turning the enterprise end of the funnel into a distribution deal.

Steal this

1. Turn your users into your content team. n8n's ~10,500 templates are user-generated SEO landing pages that double as demos. Build a dead-simple "publish what you made" flow, template the page so each one targets one high-intent query, and let the library compound. You get the SEO surface area of a 50-person content team for the cost of a submit button.

2. Make every support answer a public asset. Docs on a forum, not a Slack. Hire the community person before the revenue justifies it. Every resolved question becomes an indexed page that answers the next 500 people and pulls in organic traffic for years. Private channels burn that value; public forums bank it.

3. Re-describe your product the week the market's language changes. n8n's AI pivot was a six-week repositioning of a product they already had, and revenue 4x'd. Watch for the category rename happening in your space (it's usually visible in what your best users are already calling you), and be the first to adopt it in your homepage, your docs, and your templates. The narrative shift is cheaper and faster than any feature — and it captures the demand spike before your competitors finish their roadmap meeting.