n8n 2.0 Launched: The Self-Hosted AI Automation Stack for Technical Operators

n8n 2.0 rebuilt the visual editor from scratch and added native AI Agent nodes. Self-hosting is still free. Here is the cost math and setup reality for technical operators.

Disclosure: Some links in this article are affiliate links. We may earn a commission at no extra cost to you. See our affiliate policy.

n8n 2.0 launched in January 2026 with a rebuilt visual editor, native AI Agent nodes, and TypeScript support in the code node. The version number matters here because 2.0 is not a feature update bolted onto the existing architecture. The canvas editor was rebuilt from scratch, the AI node category is now a first-class part of the builder, and self-hosting remains free under the Fair Code license with no per-task or per-operation charges.

The audience for n8n has always been the technical operator: someone comfortable running a Docker container or a DigitalOcean droplet, who finds Zapier's per-task pricing model a poor fit for high-volume workflows. n8n 2.0 extends that audience slightly by making the UI meaningfully easier to navigate while keeping the flexibility that makes self-hosted automation worth the operational overhead in the first place.

What Changed in 2.0

The canvas editor. The original n8n editor used a flow-based layout that worked but was slow to navigate on complex workflows. The 2.0 canvas is a full rebuild: infinite canvas with pinch-to-zoom, collapsible node groups so you can fold a ten-step sequence into a single labeled block, and sticky notes for inline documentation. On a 40-node workflow, the navigation difference is significant. On a 5-node workflow, it is not.

Native AI Agent nodes. Previously, building an AI agent in n8n required custom HTTP request nodes to call OpenAI or Anthropic APIs, plus manual JSON parsing and error handling. n8n 2.0 adds a dedicated AI Agent node category with pre-built connectors for OpenAI (GPT-4o and above), Anthropic (Claude), Google Gemini, and Ollama for local model execution. The agent node handles the conversation context, tool calling loop, and response parsing. You configure which tools the agent can use (other n8n nodes, HTTP requests, or database queries) and the agent decides which ones to call based on its goal.

TypeScript in the code node. The previous code node ran JavaScript only. n8n 2.0 supports TypeScript natively, with type checking at build time. For operators writing data transformation logic with complex object structures, this removes a category of runtime errors that were previously only catchable in production.

Execution queue improvements. n8n 2.0 restructured how the execution queue handles concurrent workflows. The old architecture had bottlenecks when more than 20 workflows were running simultaneously on a single instance. The 2.0 queue supports up to 100 concurrent executions before needing to scale the worker pool. For most self-hosted single-operator stacks, this is irrelevant. For agencies running automation on behalf of multiple clients from one n8n instance, it matters.

The AI Node Architecture in Practice

The AI Agent node in n8n 2.0 works as a tool-using agent embedded in a workflow step. You give it a system prompt, a set of tools it can call, and an input. The agent runs its tool-calling loop until it produces a final output, then the workflow continues to the next node.

The tools can be anything: an HTTP request to an external API, a database query via n8n's Postgres or MySQL nodes, a file read from a local directory, or even a call to another n8n workflow via webhook. The technical comparison to Zapier Agents is instructive: Zapier Agents can use Zapier's 7,000-app integration library as tools; n8n AI Agent nodes can use anything accessible via HTTP or that n8n has a connector for. The n8n library is smaller (around 400 native integrations) but the HTTP request node covers the gap for any API that Zapier does not have a native connector for.

Practical example: an n8n 2.0 workflow that monitors an email inbox, uses an AI Agent node to classify incoming emails, queries a Postgres database to check if the sender is a known customer, and routes the email differently based on both the classification and the customer status. The entire workflow runs on your server, costs you nothing per execution beyond server electricity, and can be inspected, modified, and version-controlled as a JSON file.

Cost Math: Self-Hosted vs Cloud vs Zapier

n8n's pricing model is the main reason technical operators choose it over Zapier and Make.

Self-hosted n8n is free. The Fair Code license allows unlimited workflows, unlimited executions, and unlimited users with no per-task charges. Server cost for a DigitalOcean Basic Droplet (2 vCPUs, 2 GB RAM) is $12 per month. That covers most single-operator automation stacks comfortably.

n8n Cloud (for operators who do not want to manage infrastructure) starts at $24 per month for 5 active workflows and 5,000 executions, scaling to $50 per month for unlimited active workflows and 10,000 executions. At the $50 tier, n8n Cloud matches Zapier Professional on price but includes 5 times as many executions (10,000 vs 2,000) and does not charge per task within a multi-step workflow.

The cost comparison becomes most dramatic at high execution volumes. An operator running 20,000 monthly workflow executions on Zapier Professional ($49.99 plan) pays overage charges. On n8n Cloud Pro at $50 per month, 10,000 executions are included and additional executions are purchased in blocks. On self-hosted n8n, 20,000 executions cost nothing beyond the $12 server bill.

What Self-Hosting Actually Requires

The honest operational description: self-hosting n8n requires you to do three things that cloud automation tools do for you automatically.

First, you provision and maintain a server. A DigitalOcean Droplet, a Hetzner VPS, or an AWS EC2 instance running Ubuntu or Debian. The n8n documentation has a working Docker Compose configuration that gets you to a running instance in under 20 minutes if you are comfortable with the command line. If you are not comfortable with Linux servers and Docker, the setup time is longer and the maintenance ongoing.

Second, you handle updates. n8n releases new versions regularly. Major version updates (1.x to 2.0) occasionally require migration steps. The 2.0 migration from 1.x is documented and involved updating workflow credential storage format. It was not difficult, but it required a 30-minute maintenance window.

Third, you manage backups. n8n stores workflow definitions and credentials in a Postgres or SQLite database. If you do not back up that database, a server failure means losing your workflows. The setup is not complicated (a daily pg_dump to an S3 bucket is standard), but it is not done for you.

For technical operators already running one or two self-hosted services, n8n adds minimal overhead. For operators who have never provisioned a server, the time cost of learning is real. In that case, n8n Cloud at $24 to $50 per month is the right starting point. See the n8n vs Zapier vs Make comparison for a technical feature breakdown and the automation tools beyond Zapier and Make guide for context on when to look outside the mainstream platforms.

Who Should Use n8n vs Make vs Zapier

Use n8n if: you are technical, you run more than 5,000 workflow executions per month, you need to call arbitrary HTTP APIs that Zapier does not have native connectors for, or you need to run AI processing within workflows at a cost that does not scale with usage volume.

Use Make if: you want visual scenario building with more branching flexibility than Zapier but do not want to manage infrastructure. Make's visual builder handles complex logic more cleanly than Zapier at the same price tier, and Maia reduces the builder learning curve. At 10,000 operations per month, Make Core costs $10.59, which is dramatically cheaper than Zapier at the same volume.

Use Zapier if: you are non-technical, you want the largest app integration library, and your monthly task volume stays under 2,000. Zapier's setup is fastest, its documentation is most extensive, and its app ecosystem covers the widest range of niche tools. For simple, low-volume automations, the ease of use premium over Make and n8n is worth it.

Use the AI Stack Advisor to input your monthly execution volume, technical comfort level, and required app integrations for a platform recommendation calibrated to your actual stack.