OasisLog - 사막 같은 일상의 작은 오아시스

In the last part 1, we created the 'head' of an intelligent AI using Dify. But what can you do if you only have a head? Even if I give instructions, if the AI cannot send the email directly and does not set a schedule, I will eventually have to do it again. A true agent must be able to ‘complete’ tasks even while the user is asleep.

Today, we introduce n8n, a tool that will instill strong execution skills in AI assistants. Let's take a closer look at how to utilize this tool, which is said to be a legend in the automation market, from a veteran's perspective.

1. n8n, why is it better than paid services?

When it comes to automation, many people first think of Zapier or Make. However, as of 2026, experts choose n8n as their number one priority.

Self-hosting: Other services charge more for the more steps they automate. However, if n8n is installed on my server, the additional cost is '0 won' even if 100 steps are connected.
AI Friendly: n8n in 2026 will support LangChain nodes by default. “Is this email important?” It is easier than asking the AI to make the same judgment midway and have it perform the next task.
Visual connection: Just connect nodes with lines, like assembling Lego blocks. You can see at a glance which data is flowing and where.

2. 5-minute cut even for beginners: Installing n8n

The most stable way is to use Docker. This is the cleanest installation script recommended by an editor with 20 years of experience.

① Create a project folder

mkdir n8n-automation && cd n8n-automation

② Create `docker-compose.yml`

Copy the contents below to create a file. For beginners, we have perfectly set up data storage settings.

version: '3.8'
services:
  n8n:
image: n8nio/n8n:latest
    restart : always
    ports:
      - 5678:5678
    environment:
      -N8N_HOST=localhost
      -N8N_PROTOCOL=http
    volumes:
      - ./n8n_data:/home/node/.n8n

③ Operation!

docker compose up -d

Now you're done. Enter localhost:5678 in the browser address bar. Your own automated factory is ready to open.

3. Practice: “AI searches for news and reports on Slack”

Now let's design the entire workflow for agents to find information, summarize it, and complete the report.

1) Trigger settings: Add a 'Schedule' node to start at 9:00 every morning.

2) Think (AI Agent Node): Connect Dify agent or GPT-5.1 model through AI node inside n8n.

Prompt: “Summary today’s top tech news in 3 lines.”

3) Decision (If Node): Check whether the summary contains specific keywords (e.g. ‘Samsung Electronics’, ‘Apple’).

4) Reporting (Slack Node): Urgent news containing keywords is immediately notified via Slack on your smartphone, and the remaining information is archived in the Notion database.

4. Key details of upgrading work

Exception handling is important when creating automated workflows. Just in case the AI answer is not in the format you want, you can write a simple line of JavaScript like the one below by adding 'Code Node' to make the data much more sophisticated.

// Veteran's code to remove unnecessary whitespace from AI answers
for (const item of $input.all()) {
  item.json.output = item.json.output.trim();
}
return $input.all();

Only when these small details accumulate, 'true automation' that operates without errors in practice is completed. When structuring text, don't forget to actively utilize Markdown format to improve readability.

5. Conclusion: Automation is ‘time’, not ‘tool’.

Connecting nodes may seem complicated at first. But try to succeed just once. If you wake up and experience a report organized by AI arriving in Slack, you will never be able to go back to the past.

Let’s give repetitive tasks to machines and focus on more valuable work. In the next episode, we will cover CrewAI-based teamwork agent, which combines these tools and makes them meet each other. If you have any questions, please leave a comment anytime!