Quick Start: Your First AI Workflow

Quick Start: Your First AI Workflow#

In this notebook, you will build a working AI workflow from scratch.

No importing files. No theory first. Just building.

What you will create:

┌─────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│  Manual Trigger │────▶│   Edit Fields   │────▶│ Basic LLM Chain │
└─────────────────┘     └─────────────────┘     └─────────────────┘
                                                    ┊ (sub-node)
                                                    ┊
                                              ┌───────────┐
                                              │Chat Model │
                                              └───────────┘

This simple workflow will:

  1. Start when you click a button

  2. Create some input data (a topic)

  3. Ask an AI to write a haiku about that topic

Step 1: Create a New Workflow#

  1. Open n8n in your browser: http://localhost:5678

  2. Click Workflows in the left sidebar

  3. Click Add Workflow (or the + button)

You now see an empty canvas. This is where you build.

Tip: Click the workflow name at the top (“My workflow”) and rename it to “My First AI Workflow”.

Step 2: Add a Trigger#

Every workflow needs a starting point. This is called a Trigger.

  1. Click the + button on the canvas (or press Tab)

  2. Search for “Manual Trigger”

  3. Click it to add it to your canvas

You should see a node labeled “When clicking ‘Execute workflow’” (or similar).

What is this? The Manual Trigger starts your workflow when you click “Execute Workflow” in the editor. Perfect for learning.

Step 3: Add Input Data#

The trigger starts the workflow, but it doesn’t create any data. Let’s add a node that creates our input.

  1. Click the + button on the right side of the Manual Trigger node

  2. Search for “Edit Fields” (also called “Set”)

  3. Click it to add it

Now configure it:

  1. In the panel on the right, click Add Field → String

  2. Set Name to: topic

  3. Set Value to: artificial intelligence

Your configuration should look like:

Name

Value

topic

artificial intelligence

What is this? The Edit Fields node creates data. We’re creating a field called topic with a value. This data will flow to the next node.

Step 4: Run It and See the Output#

Let’s run what we have so far.

  1. Click Execute Workflow in the top toolbar (or press Ctrl/Cmd + Enter)

You should see green checkmarks on both nodes. Now click the Edit Fields node.

Look at the right panel. This is the Output Panel — your best friend for debugging.

Output Panel showing JSON data

You should see:

{
  "topic": "artificial intelligence"
}

The Output Panel#

Every time you run a node, you can click it to see:

  • Input — what the node received

  • Output — what the node produced

Try clicking the Table and JSON buttons to switch views. JSON shows the exact field names — you’ll need this later.

Step 5: Add the AI Node#

Now let’s add the node that calls the AI.

  1. Click the + button on the right side of the Edit Fields node

  2. Search for “Basic LLM Chain”

  3. Click it to add it

You’ll see a new node appear, but it has a warning. That’s because it needs two things:

  • A Chat Model (the AI brain)

  • A prompt (what you want the AI to do)

Step 5b: Set Up Your API Credentials#

Before connecting the AI model, you need an API key. We recommend OpenRouter — it offers free models with no strict rate limits.

Option B: Google AI Studio (Free but limited)#

  1. Go to aistudio.google.com

  2. Sign in with your Google account

  3. Click Get API Key → Create API Key

  4. Copy the API key (starts with AIza...)

Add to n8n:

  1. In n8n, click Settings → Credentials → Add Credential

  2. Search for “Google Gemini”

  3. Paste your API key → Save

Note: Google’s free tier has strict daily limits (~20-50 requests/day). If you hit rate limit errors, switch to OpenRouter.

Once saved, your credential is available across all workflows.

Step 6: Connect a Chat Model#

The Basic LLM Chain needs an AI model to work. Let’s connect one.

  1. With the Basic LLM Chain selected, look for the “Chat Model” section at the bottom

  2. Click + Chat Model

  3. Choose your provider (OpenRouter recommended)

  4. Click the model node that appeared

  5. Select the credential you created in the previous step

  6. Choose a model:

OpenRouter (recommended):

Type

Model ID

Free

deepseek/deepseek-chat-v3-0324:free

Cheap

openai/gpt-4o-mini

Powerful

anthropic/claude-sonnet-4

Google Gemini:

Type

Model ID

Free

models/gemini-flash-lite-latest

You should now see a dotted line connecting the Chat Model to the Basic LLM Chain. This means the AI brain is connected.

Step 7: Write Your Prompt#

To tell the AI what to do:

  1. Click the Basic LLM Chain node

  2. Find the Prompt field

  3. In the text area, type:

Write a haiku about artificial intelligence.

This basic approach works. To make it dynamic, use the topic field from the Edit Fields node instead of hardcoding text.

Introducing Expressions#

Expressions let you use data from previous nodes. The syntax is:

{{ $json.fieldName }}

Change your prompt to:

Write a haiku about {{ $json.topic }}.

Important: Make sure the field is in Expression mode. Click the toggle to switch from “Fixed” to “Expression”:

Expression mode toggle

What this does: When the workflow runs, {{ $json.topic }} is replaced with artificial intelligence (the value from the Edit Fields node).

Step 8: Run Your Complete Workflow#

Time for the big moment!

  1. Click Execute Workflow

  2. Wait a few seconds for the AI to respond

  3. Click the Basic LLM Chain node

In the Output panel, you should see something like:

{
  "text": "Silicon minds wake\nLearning patterns in the void\nFuture dreams take shape"
}

You did it! You just built an AI workflow from scratch.

Notice the output field is called text. This is important — when you want to use this result in another node, you would write {{ $json.text }}.

Step 9: Pin Your Data#

Every time you run the workflow, the AI generates a new haiku. But what if you want to keep working on the workflow without calling the AI again?

This is where Pinning helps.

  1. Click the Basic LLM Chain node

  2. In the Output panel, click the pin icon (📌)

  3. The node now shows a pin indicator

What this does: Next time you run the workflow, this node will use the pinned data instead of calling the AI again. This:

  • Saves money (no API call)

  • Saves time (instant result)

  • Keeps your data consistent while you build downstream nodes

To unpin: Click the pin icon again.

Pro tip: Pin data after any expensive or slow operation (like AI calls) while you’re building.

Step 10: Experiment!#

Before moving on, try these experiments:

Experiment 1: Change the topic#

  1. Click the Edit Fields node

  2. Change artificial intelligence to coffee

  3. Unpin the Basic LLM Chain (if pinned)

  4. Run the workflow again

Experiment 2: Add more fields#

  1. In Edit Fields, add another field: style = funny

  2. Update your prompt to:

    Write a {{ $json.style }} haiku about {{ $json.topic }}.

  3. Run and see the result

Experiment 3: Add a System Message#

  1. In the Basic LLM Chain, find System Message

  2. Add: You are a poet who writes in the style of Shakespeare.

  3. Run and compare the result

The more you experiment, the faster you’ll learn!

What You Learned#

In just a few minutes, you:

Concept

What you did

Workflow

Created a sequence of connected nodes

Trigger

Added a Manual Trigger to start the workflow

Edit Fields

Created input data with a topic field

Basic LLM Chain

Called an AI model with a prompt

Chat Model

Connected an AI provider (OpenRouter/OpenAI/Google)

Expression

Used {{ $json.topic }} to pass data between nodes

Output Panel

Inspected data at each step

Pinning

Saved output to avoid re-running expensive operations

These are the core skills you’ll use in every workflow.

Your Workflow Structure#

┌───────────────────────────────────┐     ┌─────────────────────────────┐     ┌─────────────────────────────┐
│  When clicking 'Execute workflow' │────▶│        Edit Fields          │────▶│      Basic LLM Chain        │
│          (Manual Trigger)         │     │                             │     │                             │
│                                   │     │  topic: "artificial         │     │  Prompt:                    │
│  Output: { }                      │     │          intelligence"      │     │  "Write a haiku about       │
│                                   │     │  style: "funny"             │     │   {{ $json.topic }}"        │
└───────────────────────────────────┘     └─────────────────────────────┘     └─────────────────────────────┘
                                                                                            ┊
                                                                                            ┊ (dotted line)
                                                                                            ┊
                                                                                    ┌───────────────┐
                                                                                    │  Chat Model   │
                                                                                    │  (OpenRouter/ │
                                                                                    │   OpenAI/     │
                                                                                    │   Gemini)     │
                                                                                    └───────────────┘

Data flow summary:

Step

Node

Output

1

Manual Trigger

{ } (empty)

2

Edit Fields

{ "topic": "artificial intelligence" }

3

Basic LLM Chain

{ "text": "Silicon minds wake..." }

Common Issues#

Problem

Solution

“No credentials” error

Go to Settings → Credentials and add your API key

Expression shows literally as {{ $json.topic }}

Switch the field from “Fixed” to “Expression” mode

Output is empty

Check the Output panel on the previous node — is the field name correct?

AI response is slow

This is normal. Pin the data while you work on other parts

Chat Model not connected

Look for the dotted line. If missing, click + Chat Model again

Recap#

In this notebook, you built your first AI workflow from scratch:

  • Created a workflow with Manual Trigger, Edit Fields, and Basic LLM Chain

  • Connected an AI provider (OpenRouter, OpenAI, or Google)

  • Used expressions ({{ $json.topic }}) to pass data between nodes

  • Learned to use the Output Panel and pinning for efficient development

Next: Proceed to 08d - Workflow Patterns to learn Core Concepts and three powerful patterns: Prompt Chaining, Routing, and Parallelization.

Contents