n8n + OpenAI Tutorial
How to Create an AI Agent in n8n Using AI Agent Node and OpenAI
Learn how to build a working AI agent in n8n using the AI Agent node, OpenAI Chat Model, tools, memory, and chat trigger.
If you want to build an AI agent inside n8n, the easiest modern setup is to use the AI Agent node together with the OpenAI Chat Model node. This lets your agent understand user requests, choose tools, and return useful answers automatically.
Important: In current n8n versions, the AI Agent now works as a Tools Agent. Also, you must connect at least one tool to the AI Agent node for it to work properly.
What You’ll Learn
- What the AI Agent node does in n8n
- How to connect OpenAI to n8n
- How to add Chat Trigger, Memory, and Tools
- How to test your first AI agent
- Common mistakes and how to fix them
What Is the AI Agent Node in n8n?
The AI Agent node in n8n is used to create an autonomous assistant that can receive input, reason about the task, and use connected tools to perform actions or retrieve information. Instead of giving only a simple text response, an agent can decide which tool to use depending on the user’s request.
This is what makes it different from a normal prompt-based AI flow. An AI agent can do more than just answer questions. It can also interact with APIs, run tools, search knowledge, and handle more advanced logic.
💡 Simple explanation: a normal AI node answers; an AI agent can also decide and act.
What You Need Before Starting
- A working n8n instance
- An OpenAI account and API key
- Basic understanding of nodes and workflow connections
n8n supports OpenAI authentication using an API key. After you add the credential, you can use it in the OpenAI Chat Model node.
Basic Workflow Structure
Your first AI agent workflow in n8n will usually look like this:
Chat Trigger → AI Agent → OpenAI Chat Model
↘ Memory
↘ Tool(s)
This structure allows the user to send a message, the AI Agent to process it, OpenAI to generate reasoning, memory to keep context, and tools to perform actions.
Step 1: Add the Chat Trigger Node
Start by adding a Chat Trigger node. This lets you test your agent using n8n’s chat interface.
- Create a new workflow
- Add Chat Trigger
- Leave the default settings for now
If you want ongoing multi-message conversations, n8n recommends connecting the same memory sub-node to both the Chat Trigger and the AI Agent. This helps maintain a single source of truth for the conversation.
Step 2: Add the AI Agent Node
Next, add the AI Agent node and connect it to the Chat Trigger.
- Add the AI Agent node
- Connect Chat Trigger → AI Agent
- Set your system prompt or instructions inside the agent
Note: Older n8n versions had multiple agent types. In current documentation, this was removed, and AI Agent now works as a Tools Agent.
Step 3: Add the OpenAI Chat Model Node
The AI Agent needs a model to think and respond. Add the OpenAI Chat Model node and connect it to the AI Agent’s model input.
- Add OpenAI Chat Model
- Create or select your OpenAI credential
- Choose the model you want to use
- Connect it to the AI Agent model connector
In the OpenAI Chat Model node, you can optionally enable Use Responses API. When enabled, it uses OpenAI’s Responses API instead of Chat Completions. n8n also notes that OpenAI built-in tools are available only when the OpenAI Chat Model is used together with the AI Agent node.
Step 4: Add at Least One Tool
This is the step many beginners miss. The current AI Agent documentation says you must connect at least one tool sub-node to the AI Agent.
Examples of tools you can connect include:
- HTTP Request tool
- Calculator tool
- Code tool
- Knowledge/search tool
- Sub-workflow tool
💡 For a simple test, add an easy tool first so your agent can actually act on something.
Step 5: Add Memory for Conversations
If you want your AI agent to remember previous messages, attach a memory sub-node. This is useful for chatbots, customer support, and assistants that need conversation context.
For chat sessions, n8n recommends connecting the same memory node to both the Chat Trigger and the AI Agent.
Step 6: Add a System Prompt
A good system prompt tells the agent what role it should play and how it should behave.
Example system prompt:
You are a helpful AI assistant for customer support.
Answer clearly and briefly.
Use connected tools when needed.
If information is missing, ask a short follow-up question.
You can customize this for sales, support, HR, booking, lead qualification, or internal knowledge assistants.
Step 7: Test the AI Agent
Once all nodes are connected, execute the workflow and test it through the chat window.
Try sample prompts like:
- “Summarize this request”
- “Use the tool to fetch data”
- “Help me answer a customer inquiry”
If everything is connected properly, the AI Agent will receive the message, use the OpenAI Chat Model for reasoning, and call the connected tool when necessary.
Simple Example Use Case
Here is a beginner-friendly example:
Chat Trigger
↓
AI Agent
├── OpenAI Chat Model
├── Simple Memory
└── HTTP Request Tool
In this setup, the user asks a question, the agent decides whether it needs external information, and the HTTP Request tool can fetch that data from an API.
Common Problems and Fixes
1. Agent does not work
Check if you connected at least one tool to the AI Agent.
2. OpenAI model not responding
Verify your OpenAI API key and selected credential.
3. Memory is not working
Make sure the memory node is connected correctly, especially if using Chat Trigger sessions.
4. Multi-item expressions behave oddly
n8n notes that sub-nodes can resolve expressions differently from standard nodes.
5. API quota or bad request errors
Check your OpenAI usage, request parameters, and model configuration.
Best Practices
- Start with one simple tool before building advanced agents
- Write a clear system prompt
- Use memory only when you actually need conversation context
- Test the OpenAI credential separately if needed
- Keep your first workflow small and easy to debug
Why This Setup Is Good
This approach is good because it is modular. You can start with a very simple AI agent, then later add more tools such as HTTP requests, Google Sheets, Telegram, CRM actions, or database queries.
It is also flexible enough to grow into customer support bots, internal assistants, data lookup bots, and AI-powered automation systems.
Watch My n8n tuts on YouTube
I also share videos about automation, n8n workflows, and AI systems on my YouTube channel.
▶ Visit My YouTube
Want More n8n Tutorials?
You can use this same setup for Telegram bots, website chatbots, CRM assistants, and AI-powered business workflows.
Read More Tutorials
FAQ
Can I build an AI chatbot with n8n and OpenAI?
Yes. The usual starter setup is Chat Trigger, AI Agent, OpenAI Chat Model, memory, and at least one tool.
Do I need a tool connected to AI Agent?
Yes. Current n8n documentation says the AI Agent requires at least one connected tool sub-node.
Can the OpenAI Chat Model use the Responses API?
Yes. The OpenAI Chat Model node has a toggle for using the Responses API.
Does memory persist forever?
n8n notes that memory does not persist between sessions unless your setup is designed to handle session continuity appropriately.
