n8n 101: Your First AI Workflow in 30 Minutes

Most developers discover n8n when they realize they have been writing the same integration glue code for the fifth time. A trigger fires, data gets fetched, an API gets called, a message gets sent. n8n handles all of that with a drag-and-drop canvas so you can focus on what actually requires your brain.

In this post you will go from zero to a working AI workflow that fetches live data, processes it with a language model, and delivers a result to a Slack channel. Thirty minutes, start to finish.

Getting n8n Running

The fastest path is n8n Cloud at n8n.io. Sign up, create a workflow, and you are on the canvas in under two minutes. If you prefer self-hosting for privacy or cost reasons, n8n runs as a Node.js process or a Docker container.

For Docker, a single command is all you need:

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

# Open http://localhost:5678 in your browser

Once you are on the canvas, you are ready to build.

The Four Core Concepts

n8n has four ideas you need to understand before the rest clicks.

Trigger. Every workflow starts with a trigger node. It can be a schedule (run every hour), a webhook (fire when an HTTP request arrives), or an app event (a new row in Google Sheets). The trigger is what wakes up your workflow.

Node. Each step in the workflow is a node. Nodes can call APIs, run JavaScript, transform data, or talk to an AI model. They are the verbs of your workflow.

Connection. Nodes are linked by arrows. Data flows along those arrows as JSON objects. Each node receives the output of the previous one and passes its own output to the next.

Execution. When a workflow runs, n8n processes each node in order, passing data through the chain. You can inspect the input and output of every node in real time.

Build: A News Summarizer Workflow

Here is the workflow you will build. A Schedule Trigger fires every morning at 8 am. An HTTP Request node fetches the latest headlines from a news RSS feed. A Code node extracts the titles and descriptions. A Claude node summarizes the top five stories. A Slack node posts the digest to your channel.

The Code node is where you do light data transformation in JavaScript. Here is the code for extracting headlines from the RSS response:

// n8n Code node: extract headlines from RSS feed items
const items = $input.all();

return items.map(item => {
  const entry = item.json;
  return {
    json: {
      title: entry.title,
      description: entry.contentSnippet || entry.content || '',
      link: entry.link,
      published: entry.isoDate
    }
  };
});

After the Code node, wire in a Claude or OpenAI node. Set the model to claude-3-5-sonnet-20241022 or gpt-4o, and use this prompt:

Testing, Debugging, and Deploying

Click the Execute button on any node to run just that node and inspect its output. This makes debugging fast: if your AI node is producing bad output, check whether the data going into it looks right before changing the prompt.

Once your workflow runs correctly end to end, click the Activate toggle in the top right. Your workflow is now live and will run automatically on the schedule you set.

One tip: always add an Error Trigger workflow that catches failures and sends you a Slack message. Silent failures are the enemy of reliable automation.