n8n AI Automations · Lesson 1

Installing n8n and your first AI workflow

Self-hosted (Docker) vs n8n.cloud, connecting OpenAI/Anthropic/Google credentials, a first HTTP Request → OpenAI → Slack workflow, and the n8n data model (items, JSON).

25 min read4 questions in quizReady prompt includedIn progress

Practical exercise

What to do after this lesson

Run n8n in Docker. Build Manual Trigger → HTTP Request (OpenAI chat/completions) → Set that puts the model's reply into an answer field. Confirm the HTTP Request preview shows one item with a choices field.

Ready-to-use prompt

Template for this lesson

Copy and adapt to your context. Text in angle brackets should be replaced.

I'm setting up self-hosted n8n in Docker for AI automations.
OS/environment: <…>
AI provider: <OpenAI / Anthropic / Gemini>

Give me: 1) docker run with volume and needed env, 2) credential creation steps,
3) a minimal HTTP Request → AI → output workflow with exact node settings.

Common mistakes

What people get wrong

Running docker run without -v volume — all workflows and credentials are lost when the container is recreated.
Pasting the API key into the JSON Body instead of using Predefined Credential Type.
Assuming a node processes all data in one call — n8n iterates per item, the AI node fires once per item.

Self-hosted vs n8n.cloud

n8n.cloud is managed hosting — no upgrades to babysit, but paid and your data flows through someone else's infra. Self-hosted (Docker) is free, full control over data (matters for PII in emails and CRM), but you handle upgrades and backups.

Minimal self-hosted launch:

docker volume create n8n_data docker run -d --name n8n -p 5678:5678 \ -v n8n_data:/home/node/.n8n \ -e GENERIC_TIMEZONE="Europe/Kyiv" \ -e N8N_HOST="n8n.example.com" \ docker.n8n.io/n8nio/n8n

The UI opens at http://localhost:5678. Workflows and credentials live in the n8n_data volume — back it up.

Connecting AI providers

Credentials → New. n8n stores keys encrypted (encryption key in N8N_ENCRYPTION_KEY); they are not exposed in workflow exports.

OpenAi credential — API Key from platform.openai.com.

Anthropic credential — API Key from console.anthropic.com.

Google Gemini (PaLM) API credential — key from Google AI Studio.

Set the key once in the credential and reuse it across all nodes. Never put the key in a URL or request body directly — only through a credential.

First workflow: HTTP Request → OpenAI → Slack

Node: Manual Trigger └─> Node: HTTP Request Method: POST URL: https://api.openai.com/v1/chat/completions Authentication: Predefined Credential Type Credential Type: OpenAi Body: JSON JSON Body: { "model": "gpt-4o-mini", "messages": [{ "role": "user", "content": "Give me an octopus fact" }] } └─> Node: Slack Resource: Message Operation: Send Channel: #general Text: ={{ $json.choices[0].message.content }}

The n8n data model: items and JSON

n8n passes an array of items between nodes. Each item is an object { json: {...}, binary: {...} }. A node runs once per item (unless configured otherwise).

{{ $json.field }} — a field of the current item.

{{ $node["HTTP Request"].json.choices[0].message.content }} — reaching into a specific node's output.

{{ $items() }} — all incoming items.

Understanding items is critical: if 10 emails come in, the AI node runs 10 times — this drives cost and rate limits. To collapse into one call, use an Aggregate or Code node before the AI node.

The most common beginner mistake is expecting a node to process "everything at once" when n8n iterates per item. Always check the item count in a node's preview.

Сообщить об ошибке

Installing n8n and your first AI workflow

Квиз — 4 вопроса

Self-hosted vs n8n.cloud

Connecting AI providers

First workflow: HTTP Request → OpenAI → Slack

The n8n data model: items and JSON