How to Connect n8n to OpenAI: Complete Integration Guide (2026)

You already know that n8n is one of the most powerful automation tools available. With over 400 built-in integrations, native AI nodes, and a fair-code license that puts you in control, it is no wonder businesses are moving away from per-execution pricing models like Zapier or Make.

But here is where things get really interesting: when you connect n8n to OpenAI, your workflows stop being simple if-this-then-that automations and start becoming intelligent. You can generate content, summarise documents, analyse customer messages, classify leads, translate emails, and even build AI agents that remember past conversations.

The best part? n8n gives you complete control. You plug in your own OpenAI API key, pay OpenAI's direct rates (no markup), and run everything from your own infrastructure.

Where you host n8n affects reliability, maintenance burden, scalability, and cost - which is why choosing the right environment matters. In this guide, we walk through everything you need to know to get your n8n and OpenAI integration up and running smoothly.

Before you begin, make sure you have:

Important: A ChatGPT Plus subscription ($20/month) does not give you API credits. The OpenAI API is billed separately on a pay-as-you-go basis. You need to add a payment method to your OpenAI account before n8n can send requests.

If you have not already, here is how to get your API key. The n8n docs outline a straightforward process:

Security tip: Store your API key securely. Never commit it to GitHub or share it in logs. In n8n, you'll store it in the credentials manager, which encrypts it automatically.

Now, let's add your API key to n8n.

The credential should now be available for any OpenAI node in your workflows. If you plan to use multiple models (for example, GPT-4o for complex tasks and GPT-4o-mini for cheaper operations), you can reuse the same credential.

There are two main approaches to integrating OpenAI into your workflows. Understanding both helps you choose the right one for your use case.

The HTTP Request node gives you complete flexibility. You can call any OpenAI endpoint - Chat Completions, Completions, Embeddings, Moderation, and more - with custom headers and payloads.

Pros: Maximum control, works with any API endpoint, no node updates required.

Cons: Requires you to build the request payload manually and parse the response.

Example payload for Chat Completions:

{
  "model": "gpt-4o-mini",
  "messages": [
    { "role": "system", "content": "You are a helpful assistant." },
    { "role": "user", "content": "Summarise this email: {{$json.email_body}}" }
  ],
  "temperature": 0.7
}

This approach is great for developers who want fine-grained control over every aspect of the API call.

n8n provides dedicated nodes that wrap the OpenAI API:

These nodes simplify configuration. You select the model, enter your prompt, and n8n handles the rest.

Pros: Faster to set up, less error-prone, automatically handles authentication.

Cons: Limited to what the node supports (newer endpoints may not be available immediately).

For most users, the built-in OpenAI Chat Model node is the easiest starting point.

Let's build a simple but useful workflow: automatic email summarisation.

That’s it. Now every time a new email arrives, you’ll get a clean summary – no more reading long threads.

For more advanced use cases - like a chatbot that remembers previous conversations or can search the web - you'll want to use n8n's AI Agent node.

The AI Agent node acts as the brain of your workflow. It orchestrates the language model, memory, and external tools to handle complex tasks.

Let's say you want a chatbot that can remember what you've already discussed and search the web for current information when needed.

Now your agent can handle questions like, What was that link you shared earlier? (memory) and What's the weather like in Tokyo today? (web search) in the same conversation.

Best practice: Consult memory first, then use tools selectively, and always summarise external results instead of returning raw search output.

Here are a few practical workflows you can build today:

These workflows can run fully automatically, saving hours of manual work every week.

All of these powerful automations depend on one thing: your n8n instance needs to be online and available around the clock. If your n8n instance goes offline, your AI workflows stop running - webhooks are missed, leads go unqualified, and support tickets pile up.

This is where your choice of hosting becomes critical.

When you're running AI agents that interact with customers or trigger based on webhooks, reliability is not optional. A self-hosted VPS might cost only $4/month, but the hidden costs - your time for setup, security patches, backup verification, and incident response - can easily reach $150-250/month.

With a managed platform like Agntable, you get:

Deploying a production-ready n8n instance with managed n8n hosting from Agntable takes minutes, not hours of YAML debugging. You can try it risk-free with a 7-day free trial.

Bottom line: If your AI workflows are important to your business or personal productivity, choosing a managed hosting option saves you time and gives you peace of mind.

Even with the right setup, things can go wrong. Here are the most common issues and how to fix them.

Why it happens: n8n cannot find your API key. This usually means the credential was not saved correctly or you are using an older node that expects an environment variable.

Fix: Go to Settings -> Credentials, check that your OpenAI credential is correctly configured and connected to the node. If you are using a community node that requires an environment variable, set it in your n8n configuration file.

Why it happens: Your OpenAI account has no remaining credits or you have hit your rate limit. A ChatGPT Plus subscription does not give you API credits - the API is billed separately.

Fix: Go to platform.openai.com/account/billing, add a payment method, and set up pay-as-you-go billing. Once that is done, n8n can send requests again.

Why it happens: You are trying to use a model name that does not exist or is not available to your account (for example, gpt-5).

Fix: Check the OpenAI models documentation for the correct model names. Use gpt-4o, gpt-4o-mini, or gpt-4.1 for current stable models.

Why it happens: The data being passed to the OpenAI node is malformed or missing required fields.

Fix: Use the Execute Workflow tab to trace the data flow through each node. Look for missing or malformed inputs before they reach the OpenAI node.

Once your workflow is working, follow these best practices to keep it reliable:

Connecting n8n to OpenAI opens up a world of intelligent automation. Whether you are summarising emails, building chatbots, or classifying leads, the combination is powerful and cost-effective.

The best part? n8n gives you complete control. You choose where to host it, you control your API keys, and you pay only for what you use.

If you are still setting up your n8n environment, check out our best n8n hosting providers guide to find the right hosting option for your needs. Already running n8n but running into Docker issues? Our n8n Docker setup guide covers the five most common failure points and how to fix them.

Now go build something intelligent.

Q: Do I need a ChatGPT Plus subscription to use OpenAI in n8n?

No. ChatGPT Plus and the OpenAI API are completely separate. The API is billed on a pay-as-you-go basis, while ChatGPT Plus is a fixed monthly subscription for using ChatGPT in the browser or mobile app.

Q: Which OpenAI model should I start with?

Start with gpt-4o-mini. It is fast, very cheap, and capable enough for most summarisation, classification, and extraction tasks. Reserve GPT-4o for complex reasoning or creative writing.

Q: Can I use other AI providers with n8n?

Yes. n8n supports Anthropic (Claude), Google Gemini, and local models via Ollama, as well as any OpenAI-compatible API.

Q: What's the difference between the HTTP Request node and the built-in OpenAI nodes?

The HTTP Request node gives you full control over the API call - you build the payload and parse the response. The built-in nodes are simpler but limited to what n8n supports out of the box.

Q: How do I handle long conversations or large documents?

Use a Memory node (like Simple Memory) to store conversation context. For large documents, consider breaking them into smaller chunks and processing them in sequence, or using a model with a larger context window (for example, gpt-4o has a 128k context window).