Why n8n Still Matters When Claude Code Can Write Your Automations

Introduction

Claude Code and Cursor can now write automation scripts from a single prompt, so it is fair to ask whether n8n, an open-source workflow automation tool, still has a reason to exist. Developers who run recurring workflows keep hearing the same question: why not generate a script and skip the visual tool? This article answers it with public adoption data collected on April 20, 2026, a look at developer tools that survived similar threats in the past, and a practical guide for choosing between them. You will get the numbers, the reasoning behind them, and a clear rule for when each tool fits.

Current Adoption: What the Numbers Show

I queried the public endpoints for Docker Hub, npm, and GitHub directly on April 20, 2026. Over the 15 days from April 5 to April 20, the n8n numbers changed as follows.

Metric	2026-04-05	2026-04-20	15-day increase
Docker pulls (n8nio/n8n)	199,717,638	204,452,618	+4,734,980 (+2.4%)
GitHub Stars	182,400	184,771	+2,371
npm weekly downloads	-	71,835	(monthly 314,419)

Docker pulls grew by 4,734,980 in that window, a 2.4% increase. GitHub stars rose by 2,371 over the same period. On April 20, the npm weekly download count for n8n stood at 71,835, with a monthly count of 314,419.

Those figures only make sense next to the numbers for adjacent tools. The table below places n8n alongside workflow engines, AI agent frameworks, and an AI coding CLI, all measured on the same date.

Tool	Docker Pulls	npm weekly DL	GitHub Stars	Category
n8n	~204 million	71,835	184,771	Workflow automation (open source, GUI)
Apache Airflow	1.577 billion	—	45,100	Pipeline orchestration (code-first)
Temporal	41.55 million	1,411,156	19,702	Workflow engine (code-first)
LangGraph	—	2,056,118	29,694	AI agent framework (code-first)
Grafana (reference)	5.192 billion	—	—	Monitoring
Metabase (reference)	253 million	—	—	BI
Portainer (reference)	1.442 billion	—	—	Docker GUI
Zapier	—	—	—	SaaS (closed; metrics not public)
Make	—	—	—	SaaS (closed; metrics not public)
Claude Code	—	13,147,573	116,032	AI coding CLI
@anthropic-ai/sdk	—	14,792,212	—	Anthropic API SDK
Cursor	—	—	Private (private repository)	AI IDE

Three observations stand out.

First, Claude Code’s weekly npm downloads reached 13,147,573. That is about 183 times n8n’s 71,835 weekly downloads. Claude Code is spreading fast across developer machines.

Second, n8n’s Docker pulls still grew by 4,734,980 in two weeks, even while Claude Code was expanding. Deployment momentum did not slow down.

Third, the two metrics count different things. A Claude Code install happens once per developer machine, and developers reinstall or update it weekly. An n8n Docker pull reflects a server deployment, and the note reads the 204 million cumulative pulls as tens to hundreds of thousands of always-on instances. In other words, Claude Code measures how often developers use a tool, while n8n measures how many running systems depend on it. Comparing them directly overstates the gap.

Historical Precedent: Tools That Survived the Same Threat

This is not the first time a new approach has seemed likely to absorb an established tool. Several long-lived tools faced similar threats and kept their place.

Tool	Threat	Years survived	Why it survived
Excel	Tableau, Power BI, Looker	40+	Flexibility, zero entry barrier, local execution
Git	GitHub Desktop and other GUI tools	20+	Standard CLI, scriptable
VS Code	Cursor, Windsurf	—	(Instead, it was absorbed as the underlying engine)
SQL	NoSQL, ORMs, GraphQL	50+	The lingua franca of data languages
Bash/Shell	Python and Node scripts	40+	Local and offline, pipe standard

Excel is the clearest case. Tableau, Power BI, and Looker arrived, yet Excel remains the world’s largest BI tool. Git shows the same pattern: GUI clients appeared, but developers still type git in the terminal. SQL persisted through the NoSQL, ORM, and GraphQL waves. Bash continues to serve as the default for system automation even as Python and Node scripts spread.

The note extracts three shared traits from these survivors:

1. They run locally or offline. They do not depend on external APIs or network access.
2. They expose a standard input and output interface. JSON, plain text, and pipes make them interoperable.
3. They have a community plugin ecosystem. Community contributions meet long-tail demand that a central vendor cannot cover.

Applying these criteria to n8n gives a favorable result. n8n can be self-hosted and run locally, it uses a JSON-based workflow format, and it offers more than 1,000 community nodes and over 900 official templates. By the note’s reasoning, n8n meets all three conditions.

The Three Moats n8n Defends

The note argues that n8n survives the vibe coding wave because of three structural advantages. Each one addresses a specific weakness of AI coding tools in production settings.

Data Sovereignty Through Self-Hosting

Claude Code works by sending API calls to Anthropic’s servers. That is convenient, but sending customer data to an external AI provider can rule the tool out in regulated industries. The note lists three examples:

• Financial sector: the Electronic Financial Supervision Regulations (South Korea’s financial-sector rules) restrict transferring customer information overseas.
• Healthcare: the Personal Information Protection Act (South Korea’s data privacy law) and HIPAA (the US health data law) limit how patient data can be processed.
• European business: GDPR requires data subject consent and clear statements about where data is processed.

n8n can run inside a company network, on premises, or in Docker. Data stays inside the internal network. In regulated industries, the note argues, this is not a matter of preference. It determines whether a team is allowed to adopt the tool at all. For these organizations, n8n is effectively the only workflow engine they can choose.

Speed of Prototyping AI Agents

The note describes building an AI agent by dragging three nodes onto a canvas: an HTTP node, then an OpenAI or Claude node, then a Slack node. The agent can be running in about 30 minutes. Triggers, scheduling, retries, and log retention come with the platform.

Claude Code generates code quickly, but running that code continuously requires adding scheduling, error handling, log management, and retry logic by hand. The note calls this surrounding boilerplate the main cost of moving from prototype to production. In its framing, n8n acts as the runtime shell for AI agents: the stage where code actually runs over time.

Free Licensing and Lower Cost at Scale

Teams can self-host n8n without a license fee, and the note describes the Fair-Code license as making unlimited self-hosted use free. Claude Code uses token-based pricing, so a workflow that runs 1,000 times accumulates charges with each run. The note also cites earlier internal research showing that paid n8n Cloud plans cost roughly 60% less than Zapier. That research is referenced in the note but not reproduced in it.

The note’s conclusion on cost is specific. One-off tasks are easier with Claude Code. Repeatedly operated pipelines are cheaper to run on n8n when measured by total cost of ownership.

Practical Decision Guide

The choice depends on the situation. The note’s recommendations are summarized below.

Situation	Recommended tool
One-off script, local exploration	Claude Code
Running workflow, shared by a team	n8n
Handing off to non-developers, GUI needed	n8n (the GUI serves as documentation)
Self-hosting required (regulated industry)	n8n
Building a custom node	Generate code with Claude Code, then run it in n8n

The note does not frame these tools as competitors. It frames them as layers. I generate n8n custom node code with Claude Code and then run the nodes in n8n. The distinction between the tool that generates code and the runtime where code lives over time fits real work best, and it explains why the two tools coexist without one replacing the other.

Bottom Line

Each time a new paradigm arrives, someone declares an older tool dead. The history of Excel, Git, and SQL suggests otherwise. Tools that run locally, expose standard interfaces, and have deep community ecosystems rarely disappear. The April 2026 data supports the same conclusion for n8n. Its Docker pulls kept climbing during the rise of Claude Code, and its strongest advantages (self-hosting, a runtime for agents, and a cost structure suited to recurring work) are the ones that matter most in production. The evidence points to layer separation rather than replacement: AI coding agents handle prototyping, and workflow engines like n8n handle production operations.

Sources

• Docker Hub API (n8nio/n8n): https://hub.docker.com/v2/repositories/n8nio/n8n/
• npm registry (n8n weekly downloads): https://api.npmjs.org/downloads/point/last-week/n8n
• GitHub REST API (n8n repository): https://api.github.com/repos/n8n-io/n8n
• Stack Overflow Developer Survey, yearly Excel and SQL usage rates (cited in the note; no URL recorded)
• GitHub Octoverse (cited in the note; no URL recorded)
• Earlier internal research: 2026-04-05-n8n-market-analysis.md