N8n

From PRIMO.ai
Jump to navigation Jump to search

YouTube ... Quora ...Google search ...Google News ...Bing News

n8n (pronounced "n-eight-n") is a workflow automation platform that lets people connect apps, APIs, databases, and large language models on a visual canvas, then run the resulting workflows on a schedule, from a webhook, or in response to a chat message. It was created in Berlin by Jan Oberhauser, who spent more than a decade in software engineering and the visual effects industry before launching the project publicly in October 2019, and the company behind it, n8n GmbH, was registered the following month. The name is a numeronym for "nodemation," a blend of node, Node.js, and automation, which reflects how every step in a workflow is a node and how the product itself is built on Node.js and TypeScript. Unlike most automation services, n8n can be run on your own servers through a free self-hosted Community Edition or used as a managed service called n8n Cloud, and builders can drop into JavaScript or Python Code nodes whenever a visual step is not flexible enough. Over the past two years it has become one of the most popular places to build AI agents, combining LangChain-based agent nodes, memory, tools, and Model Context Protocol (MCP) support with more than a thousand prebuilt integrations.

The platform matters because most useful AI work is not a single chat window but a chain of steps that touch email, spreadsheets, CRMs, ticketing systems, and internal databases, and n8n gives those steps a visible, auditable shape. Think of it like the wiring closet of an office building: the AI model is the smart thermostat, but someone still has to run cables to every room, label them, and install breakers so a fault in one place does not take down everything else. n8n has grown quickly by filling that role, passing $100 million in annual recurring revenue (ARR) by mid-2026, reaching a $2.5 billion valuation in its October 2025 Series C, and roughly doubling that to about $5.2 billion when SAP made a strategic investment in May 2026. At its first conference in October 2026, the company reported 6.7 million active workflows a month, use at 60% of the Fortune 500, and a community of more than 500,000 members.


n8n turns AI from a conversation into infrastructure, giving builders a visual, self-hostable control plane where models, tools, data, and human approvals are wired together into workflows that run reliably in production.


History and Company

Oberhauser built the first version of n8n as a side project, drawing on his experience building production pipelines in the visual effects industry, where many specialized tools and systems have to be connected to one another. He announced it on Hacker News in October 2019 under the Apache 2.0 license with the Commons Clause, a combination that allowed free use and modification but prohibited selling the software itself as a service. In March 2020 the company raised a $1.5 million seed round from Sequoia and firstminute capital, followed by a $12 million Series A led by Felicis in April 2021. Early adopters were largely developers who wanted a self-hostable, extensible alternative to cloud automation services such as Zapier, and that technical community remains the core of n8n's audience.

In March 2022 n8n replaced its original license with the Sustainable Use License, which it created and describes as "fair-code." The license lets anyone use, modify, and redistribute the software for internal business purposes or personal use, while requiring a commercial agreement for products or services whose value comes entirely or substantially from n8n functionality; consulting and building workflows for clients remain allowed. Because the Open Source Initiative does not accept licenses with limits on use, n8n does not call itself open source, and some developers criticized the company for blurring that line, while others saw fair-code as an honest middle ground. Oberhauser has since said that trust and consistency matter more to a community than features, and he credits that approach with helping n8n avoid the community backlash that hit other open-source companies over licensing.

The arrival of generative AI turned n8n from a niche developer tool into one of Europe's fastest-growing software companies. After a €55 million Series B led by Highland Europe and HV Capital in March 2025, n8n raised $180 million in a Series C led by Accel in October 2025 at a $2.5 billion valuation, with participation from Meritech, Redpoint, Visionaries Club, Evantic, T.Capital, and NVIDIA's venture arm NVentures, bringing total funding to about $240 million. On May 12, 2026, SAP announced a strategic investment through a secondary share purchase that valued n8n at about $5.2 billion, together with a multi-year agreement to embed n8n in Joule Studio, part of SAP's Business AI platform. At the time of that deal n8n reported more than 1,400 enterprise customers and 1.7 million monthly active builders, and on GitHub it had added more than 112,000 stars in 2025 alone to top the JavaScript Rising Stars ranking before passing 200,000 stars in 2026.

In October 2026 n8n held In The Loop, its first conference, at Wilhelm Studios in Berlin on October 13 and 14, bringing builders, partners, and enterprise customers to its hometown. In his opening keynote Oberhauser argued that more capable AI should make people more capable, and the company used the event to unveil a refreshed brand with a new logo, typeface, and color system, along with n8n Labs for experimental features. Product leaders previewed simpler evaluations, a new execution engine, stronger governance and guardrails, and an observability product called Manifest that was in testing, and the company pledged to keep half of its engineers focused on the core workflow product. Public reference customers named in 2026 include Vodafone, Mercedes-Benz, Accenture, Dell, and Deutsche Telekom, which also became a certified n8n partner.

How n8n Works

Every n8n automation is a workflow: a directed graph of nodes on a canvas, where data flows from one node to the next as a list of JSON items. A workflow starts with a trigger node, such as a schedule, a webhook, a form submission, a new email, a row added to Google Sheets, or a chat message, and continues through action nodes that call services such as Slack, Salesforce, Notion, GitHub, or a database. Between them sit logic and utility nodes (If, Switch, Merge, Loop Over Items, Wait, and Edit Fields) plus Code nodes, and expressions written in double curly braces, such as {{ $json.email }}, let any field reference data produced by earlier steps. When a service has no built-in node, the HTTP Request node can call almost any REST API, and it can be configured by importing a cURL command copied from the service's documentation.

While you build, each node can be run on its own and shows the exact input and output data it handled, which makes debugging feel closer to stepping through a spreadsheet than reading server logs. n8n 2.0 introduced a Save versus Publish model, so edits can be saved as drafts without changing the version that is live in production. Each complete run is recorded as an execution, and n8n Cloud pricing is based on executions rather than individual steps, so a workflow with twenty nodes costs the same per run as one with two. For larger deployments, queue mode separates the main instance from worker processes that pull jobs from Redis, and in n8n 3.0 even manual test runs execute on workers so that development activity cannot slow production.

n8n can run as a managed service on n8n Cloud or be self-hosted in a Docker container, on Kubernetes, or on a virtual machine from almost any hosting provider. Self-hosting appeals to organizations with strict data governance or privacy requirements, because workflow data, credentials, and execution history stay on infrastructure they control, and the same instance can call local models running on their own hardware through nodes such as Ollama. Since n8n 2.0, Code nodes run in separate task runners by default, isolating custom code from the main process, and access to environment variables from workflows is restricted. Business and Enterprise customers can also store workflows in Git and promote them between development, staging, and production environments, which brings automation closer to normal software development practice.

AI Agents in n8n

n8n's AI features have grown from a single experimental node into a full agent-building platform, and they are now the main reason many new users arrive. They fall into four layers: the AI Agent node for adding reasoning inside a workflow, n8n Agents for standalone agents that live outside any single workflow, the AI Assistant that builds workflows from plain-language requests, and Model Context Protocol (MCP) support that connects n8n to coding agents and other AI clients, with supporting features for retrieval, testing, guardrails, and chat around them. Each piece is described below, and readers comparing n8n with other tools can consult the Visual and Low-Code Agent Builder Comparison and the Agent Development Framework Comparison.

AI Agent Node

The AI Agent node is built on LangChain cluster nodes, meaning the agent is a root node with sub-nodes attached beneath it for a chat model, memory, and tools. The chat model can come from OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek, OpenRouter, Ollama for local models, and other providers, and memory options range from a simple in-workflow buffer to Postgres, Redis, and other persistent stores. Tools can be almost any n8n integration, another workflow, an HTTP request, a code snippet, a vector store for retrieval-augmented generation, or an MCP Client Tool that reaches external MCP servers, and the agent chooses which tool to call based on the user's request and each tool's description. Since version 2.6 in January 2026, individual tool calls can require human approval through channels such as Slack, email, or chat before they run, which makes it practical to let agents draft refunds, send messages, or change records under supervision.

Evaluations, RAG, and Guardrails

n8n also ships the building blocks for retrieval-augmented generation, including document loaders, text splitters, embedding models, and vector store nodes for services such as Pinecone, Qdrant, Supabase, and Postgres with pgvector, plus a simple in-memory store for experiments. Since mid-2025, Evaluations let builders run an AI workflow against a dataset of test inputs and score the results with metrics such as correctness or helpfulness, so changes to prompts and models can be measured rather than guessed. Structured output parsers, a Guardrails node for screening inputs and outputs, and fallback models that take over when a primary model fails round out the toolkit, and Oberhauser has demonstrated agents switching from Claude to GPT through OpenRouter when a model is unavailable. At In The Loop 2026 the company said simpler evaluations and stronger governance features were next on its roadmap, reflecting how enterprise buyers now ask about testing and control before they ask about features.

n8n Agents

In August 2026 n8n introduced n8n Agents, a new kind of object that lives alongside workflows rather than inside them. An n8n Agent bundles a model, instructions, tools (integrations, workflows, JSON-schema tools, and MCP servers), reusable skills, knowledge files, memory, approval rules, schedules, and sub-agents into one configuration that is set up once and reached from many places. Agents can be connected to channels such as Slack, Telegram, and Linear, called from any workflow through a Message an Agent node, versioned with the same draft and publish model as workflows, and reviewed through a Sessions tab that records each conversation. The feature launched as a preview on n8n Cloud and a beta for self-hosted instances running version 2.32.3 or later, n8n highlighted Agents as shipping in October during its In The Loop keynote, and each agent turn counts as one execution for billing.

Chat Hub

Chat Hub, introduced in early 2026, gives teams a chat interface inside n8n for talking to models and to agents built on the instance, so colleagues can use AI without opening the workflow canvas. n8n 3.0 disables Chat Hub by default, so organizations that depend on it need to switch it back on after upgrading. For reaching people where they already work, n8n now also offers n8n Agents connected to channels such as Slack and Telegram, which are described above.

AI Assistant

The AI Assistant lets users describe an automation in plain language, such as "every morning, summarize unread customer emails and post the summary to Slack," and then builds, explains, or debugs the corresponding workflow on the canvas. It became available on both n8n Cloud and self-hosted instances in July 2026 with version 2.29.9, and n8n has reported about 1.6 million assistant conversations and roughly 800,000 workflows built with it. Usage draws on monthly AI Builder credits included with each paid plan, currently 50 on Starter, 150 on Pro, and 1,000 on Enterprise, and n8n 3.0 removes the older Ask AI tab inside the Code node. A related convenience feature, n8n Connect, sets up credentials automatically for supported services and is already used by about a quarter of users, removing one of the most common points where beginners get stuck.

Model Context Protocol (MCP)

n8n supports the Model Context Protocol in three ways, each solving a different integration problem. The MCP Server Trigger node turns any workflow into an MCP server whose tools external AI clients can call, and the MCP Client Tool lets an AI Agent node use tools exposed by other MCP servers, with n8n advertising simpler connections to about 70 popular MCP servers. The third, an instance-level MCP server, is enabled under Settings and secured with OAuth or an access token, and it lets coding agents such as Claude Code, OpenAI Codex, and Cursor find, run, create, and edit workflows on an n8n instance; workflow building was added to it in version 2.14.0. This turns n8n into a shared automation backend for developers who prefer to work from a terminal or IDE, while non-technical colleagues continue to see and adjust the same workflows visually.

n8n Products and Offerings

Because n8n is a company with several distinct offerings rather than a single app, the table below summarizes each product and AI capability as of late 2026. Everything runs on the same core workflow engine, so a workflow built on the free Community Edition can usually be moved to n8n Cloud or an Enterprise deployment by exporting and importing its JSON definition. The main differences between editions are hosting, collaboration, governance, and support, although some newer AI services, such as the AI Assistant, depend on plan credits.

Offering What it is Notes
Community Edition Free, self-hosted n8n under the Sustainable Use License Unlimited workflows and executions; community support; Docker is the only supported install method from n8n 3.0
n8n Cloud Managed hosting operated by n8n Starter, Pro, and Enterprise tiers priced by executions; 14-day free trial
Business and Enterprise Editions with single sign-on, Git-based environments, role-based access, and advanced administration Business is self-hosted; Enterprise adds custom limits, 200 or more concurrent executions, and dedicated support
n8n Embed A commercial license for building n8n's automation features into another company's product Required for uses the Sustainable Use License does not allow
AI Agent node LangChain-based agent with chat model, memory, and tool sub-nodes Available in every edition; supports human approval of individual tool calls
n8n Agents Standalone, reusable agents with skills, knowledge, channels, sub-agents, approvals, and schedules Preview from August 2026 on version 2.32.3 or later; one agent turn equals one execution
AI Assistant Natural-language workflow builder, explainer, and debugger Uses monthly AI Builder credits; available on Cloud and self-hosted since July 2026
Evaluations Dataset-driven testing and scoring for AI workflows Introduced in mid-2025; simplified evaluations announced for the roadmap in October 2026
MCP support MCP Server Trigger, MCP Client Tool, and an instance-level MCP server The instance-level server lets coding agents build and edit workflows
Chat Hub A chat interface for talking to models and agents inside n8n Introduced in early 2026; disabled by default in n8n 3.0
Templates and integrations Official and community workflow templates plus built-in and community nodes About 13,000 templates and 2,300 integrations cited in October 2026; unverified community packages are off by default in n8n 3.0
n8n Labs A home for experimental features Announced at In The Loop 2026 alongside the brand refresh

Pricing

n8n Cloud charges by executions, meaning complete runs of a workflow, rather than by the number of steps or tasks inside each run, which makes long multi-step automations much cheaper than on per-task platforms. The self-hosted Community Edition is free with unlimited executions, but the owner pays for servers and is responsible for maintenance, backups, scaling, and security updates. Prices below are n8n's published euro rates in late 2026; annual billing is about 17% cheaper than monthly billing, cloud plans start with a 14-day free trial, and companies with fewer than 20 employees can apply for a start-up plan at 50% off Business.

Plan Price Executions Highlights
Community Edition Free Unlimited Self-hosted; all core nodes and AI nodes; community support
Starter €20 per month billed annually (€24 monthly) 2,500 per month n8n Cloud; 5 concurrent executions; 50 AI Builder credits
Pro €50 per month billed annually (€60 monthly) 10,000 per month n8n Cloud; 20 concurrent executions; 150 AI Builder credits
Business €667 per month billed annually (€800 monthly) 40,000 per month Self-hosted; single sign-on, Git version control, and environments; extra executions sold in blocks of 300,000 for €4,000
Enterprise Custom Custom Cloud or self-hosted; 200 or more concurrent executions; 1,000 AI Builder credits; dedicated support and governance controls

Security

n8n's popularity and its position at the center of company systems have made it an attractive target, because a single instance often holds credentials for email, CRMs, databases, and cloud accounts. In January 2026 researchers at Cyera disclosed Ni8mare (CVE-2026-21858, CVSS 10.0), a Content-Type confusion flaw in webhook and form handling that let unauthenticated attackers read files from servers running versions 1.65.0 through 1.120.x. n8n had already fixed the flaw in version 1.121.0 on November 18, 2025, but Cyera estimated that about 100,000 servers were potentially affected and Censys counted roughly 26,500 exposed hosts when the details became public. Around the same time n8n patched CVE-2026-21877 (CVSS 10.0, a file upload flaw fixed in 1.121.3), CVE-2025-68613 (CVSS 9.9, remote code execution fixed in 1.120.4, 1.121.1, and 1.122.0), and the "N8scape" sandbox bypass CVE-2025-68668, fixed in 2.0.0.

The pace of fixes continued through 2026, which is typical for widely deployed software but demanding for self-hosted administrators. An October 14 bulletin covered eight vulnerabilities, five rated high, fixed in versions 1.123.85, 2.43.5, and 2.44.1, and a further security update on October 28 prompted Canadian Centre for Cyber Security advisory AV26-1103, with releases 2.46.3, 2.45.7, and 1.123.91 following on October 30 to enforce the execution:read permission. n8n 3.0 tightens several defaults in response, including always-on encryption key rotation, a larger server-side request forgery (SSRF) block list, mandatory signing secrets for Slack credentials, and disabling unverified community packages. Good security practice is to keep instances updated, avoid exposing the editor directly to the internet, restrict who can create webhooks and install community nodes, and treat agents with tool access as prompt injection targets, since a malicious email or web page read by an agent could try to steer its tool calls.

Getting Started: Build Your First AI Agent

The fastest way to learn n8n is to build a small AI agent that chats, remembers what you said, and uses a tool, because that one workflow exercises triggers, the AI Agent node, memory, and tool calling. The example below takes about 15 minutes on n8n Cloud and needs no code, and it mirrors the structure of n8n's own introductory AI tutorial. Start a 14-day free trial at n8n.io, which gives you a hosted instance and a small allowance of free OpenAI API credits for experimenting, or self-host on your own computer with Docker using the two commands below. Docker becomes the only supported self-hosting method in n8n 3.0, and once the container is running you open http://localhost:5678 in a browser and create the owner account. 

docker volume create n8n_data
docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n

With an instance ready, build the agent in seven steps. Node names below match the current editor, although menu labels occasionally shift between releases. Save the workflow after each step so nothing is lost if the browser tab closes.

  1. Create a workflow. From the Overview page, create a new workflow and name it My First Agent.
  2. Add a chat trigger. Select Add first step, search for Chat Trigger, and add it; the node appears as When chat message received and gives you a built-in chat panel for testing.
  3. Add the agent. Select the plus (+) connector on the trigger, search for AI Agent, and add it, leaving the prompt source set to the connected chat trigger.
  4. Attach a chat model. Select the plus under the agent's Chat Model connector, choose OpenAI Chat Model, and create a credential using the free n8n credits or your own OpenAI API key, then pick a small, inexpensive model; Anthropic, Gemini, and Ollama models work the same way.
  5. Add memory. Select the Memory connector and choose Simple Memory; because the chat trigger is connected, the session key is taken from {{ $json.sessionId }} automatically, and the default context window is enough for testing.
  6. Give it a tool. Select the Tool connector and add the Calculator tool, which lets the model hand arithmetic to a deterministic calculator instead of guessing.
  7. Write instructions. Open the AI Agent node, add a System Message under Options such as the example below, and save the workflow.
You are a friendly assistant for a small business.
Use the Calculator tool for any arithmetic and show the result clearly.
If you do not know something, say so instead of guessing.

To test the agent, select the Chat button near the bottom of the canvas and send the three messages below one at a time. The first checks tool use: the agent should call the Calculator and reply that 17.5% of 2,480 is 434, and the Calculator sub-node will light up in the execution view so you can see exactly what was sent and returned. The last two check memory, since the agent should answer "Sam" only because Simple Memory carried your earlier message forward. If anything fails, open the node marked with a warning, read the error in its output panel, and fix the credential or connection before chatting again. 

What is 17.5% of 2,480?
My name is Sam.
What's my name?

Once the basic agent works, the natural next steps are to make it useful and shareable. Swap the chat trigger for a Gmail, Slack, or Telegram trigger so the agent responds where people already work, add tools such as Google Sheets or an HTTP Request to a company API, and connect a vector store if it needs to answer from your own documents. To share it, publish the workflow and make the chat trigger publicly available, which gives you a hosted chat page or an embeddable widget, and add human approval to any tool that sends messages or changes records. Before connecting real customer data, review the Security section above and build a small evaluation dataset so you can see how changes to prompts or models affect results, and browse the ...build an agent page for broader agent design guidance.

Why This Matters

n8n sits at an important junction in the AI market, between the model makers that supply intelligence and the business systems where work actually happens. For operations, marketing, finance, and IT teams, it offers a way to turn AI from a chat tool that individuals use privately into shared, documented automations that run every day, with every step visible on a canvas and every run logged. Oberhauser has argued that AI sprinkled on top of a product rarely works, that AI belongs inside the core workflow, and that the people who own a problem should build its automation instead of waiting for a central AI team, and n8n's design follows from that view by letting a domain expert start visually and a developer extend the same workflow with code. Compared with Zapier and similar services, n8n trades some polish for control through self-hosting, code, and execution-based pricing, and compared with traditional robotic process automation, it works mainly through APIs rather than by imitating clicks on screens.

The trade-offs are real and worth understanding before a large rollout. Self-hosting shifts responsibility for patching, backups, scaling, and security onto the owner, and the 2025 and 2026 vulnerabilities showed how costly that responsibility can be when an internet-facing instance falls behind on updates. The fair-code license gives most organizations what they need, but it is not an OSI-approved open source license, so companies that resell automation built substantially on n8n need a commercial agreement. Agents built in n8n also inherit the usual weaknesses of LLMs, including hallucination, inconsistent tool selection, and exposure to prompt injection, which is why approvals, evaluations, and narrowly scoped credentials matter more than clever prompts.

The SAP partnership points to where the platform is heading. By embedding n8n in Joule Studio, SAP is betting that enterprises want an open orchestration layer that reaches systems outside SAP as well as inside it, and n8n's environments, role-based controls, and planned guardrails are increasingly aimed at that audience and at the compliance teams that review it. For readers deciding where n8n fits among agent tools, the AI Agent Environment Selector and the Ready-to-Use AI Agent Comparison place it alongside code-first frameworks, hosted assistants, and other visual builders.

Recent Developments

Between 2025 and late 2026, n8n evolved from a developer-favorite automation tool into an enterprise AI orchestration platform, shipping two major versions, launching standalone agents, and drawing a strategic investment from SAP. Most of these changes are documented on the n8n blog, in its community forum, and in release notes that appear almost weekly. The pace has been fast enough that tutorials more than a year old often show a noticeably different interface. The most significant developments were:

  • Series B and Series C (March and October 2025): n8n raised a €55 million Series B led by Highland Europe and HV Capital in March 2025, then a $180 million Series C led by Accel in October 2025 at a $2.5 billion valuation, with NVIDIA's NVentures among the participants. Its ARR was reported above $40 million during 2025, and in August 2025 the European tech publication Sifted questioned whether the speed of its valuation increases was justified.
  • Evaluations and AI Building Blocks (mid to late 2025): n8n added Evaluations for testing AI workflows against datasets, along with guardrail and fallback-model features that made agents easier to run in production.
  • n8n 2.0 (December 2025): The 2.0 beta arrived on December 8 and the stable release on December 15, 2025. It introduced the Save versus Publish model, task runners that isolate Code nodes by default, restricted environment-variable access, a Migration Report for upgrades, a pooled SQLite driver that n8n said was up to 10 times faster, and improvements to the Wait node and human-in-the-loop sub-workflows, while 1.x received three months of security support.
  • Ni8mare and a Wave of Critical Vulnerabilities (January 2026): Cyera's January 7 disclosure of CVE-2026-21858 drew wide attention to unpatched self-hosted instances, alongside several other critical CVEs patched in late 2025.
  • Human Approvals and Chat Hub (January 2026): Version 2.6 added human approval for individual agent tool calls, and n8n introduced Chat Hub as a chat interface for models and agents inside the product.
  • SAP Strategic Investment (May 2026): SAP bought shares in a secondary sale that valued n8n at about $5.2 billion and signed a multi-year agreement to embed n8n in SAP Joule Studio, while Deutsche Telekom became a certified partner during the year.
  • AI Assistant and $100 Million ARR (July 2026): The AI Assistant became available on Cloud and self-hosted instances, and Oberhauser said on The Product Podcast that n8n had passed $100 million in ARR.
  • n8n Agents (August 2026): n8n introduced standalone Agents with skills, knowledge, channels, approvals, sub-agents, schedules, and a Message an Agent node, first as a Cloud preview and self-hosted beta.
  • In The Loop 2026 and Brand Refresh (October 13 to 14, 2026): At its first conference in Berlin, n8n reported 6.7 million active workflows a month, use at 60% of the Fortune 500, 519,000 community members, 13,000 templates, and 2,300 integrations. It unveiled a new brand and n8n Labs, said Agents were shipping, previewed simpler evaluations, a new execution engine, governance and guardrails, and the Manifest observability tool, and pledged to keep half of its engineers on the core product.
  • n8n 3.0 Beta (October 27, 2026): The 3.0 beta removes legacy nodes (Function, Function Item, Item Lists, Cron, Interval, the legacy OpenAI node, and older AI Agent modes), Import from URL, and the Code node's Ask AI tab, and disables Chat Hub by default. Manual executions now run on workers in queue mode, in-memory binary data mode is gone, the Code node timeout drops from five minutes to one, and self-hosting requires Docker instead of npm or npx, with a one-line setup script offered. Stable release is expected in early December 2026, 2.x will receive 12 months of fixes, 1.x support ends on December 31, 2026, and the move to 3.x is one-way.
  • Ongoing Security Updates (October 2026): Security bulletins on October 14 and October 28, a Canadian Centre for Cyber Security advisory, and patched releases on October 30 kept vulnerability management at the top of administrators' agendas.
  • Debates Over Valuation, Licensing, and Production Readiness: Supporters point to n8n's revenue growth, enterprise adoption, and SAP's endorsement as evidence that open, self-hostable orchestration is becoming core AI infrastructure. Skeptics have questioned how quickly its valuation rose, whether fair-code should be marketed alongside open source, and whether citizen-built agents can be governed safely at scale. The string of critical vulnerabilities and the breaking changes in 3.0 have also reminded teams that running n8n in production requires the same discipline as any other business-critical software.

Additional Viewing

Load video
YouTube

In this July 2026 episode of The Product Podcast from Product School, n8n founder and CEO Jan Oberhauser explains how the company passed $100 million in annual recurring revenue, why its fair-code license helped it avoid the community backlash that hurt other open-source companies, and why AI should be built into the core workflow rather than sprinkled on top. He also discusses agent guardrails that limit each agent to explicitly permitted actions, argues that the people closest to a problem should build its automations, and demonstrates a personal email and calendar agent that falls back from Claude to GPT through OpenRouter. It is the best single source for the strategy behind the History and Company and Why This Matters sections.

Load video
YouTube

This roughly 90-minute masterclass from Nate Herk of the AI Automation channel walks beginners from the n8n interface and core node types through building working AI agents with chat models, memory, and tools. It finishes with best practices and suggested next steps, making it a natural follow-up to the Getting Started example on this page. Because n8n's interface changes often, some menus may look slightly different in current versions.

Load video
YouTube

Published by AI Foundations in June 2025, this full course explains what AI agents are, then teaches n8n's node types, including triggers, actions, utilities, code, and agent nodes, before building complete agents step by step. It is aimed at non-technical viewers and encourages them to apply knowledge of their own industry when choosing automations worth building. It is a useful companion to the How n8n Works and AI Agents in n8n sections.

Load video
YouTube

This roughly four-hour freeCodeCamp course, developed by Marconi Darmawan of KodeKloud and announced in December 2025, takes viewers from n8n basics to production-grade agents. Projects include a research workflow using Perplexity and OpenAI, a customer support RAG agent backed by a vector database, Slack and multimedia agents, and a multi-workflow team of agents that replaces a manual process. It is the most thorough free option for readers who want to go well beyond the beginner example on this page.

References

Retrieved from "https://primo.ai/w/index.php?title=N8n&oldid=52587"