OpenAI API Key: How to Get One and Use It Safely

To get an OpenAI API key, sign in at platform.openai.com, open the API keys page in your organization settings and create a new secret key. Then add prepaid credits (the minimum purchase is $5), store the key in an environment variable named OPENAI_API_KEY, and use it only from server-side code.

This guide follows OpenAI's own pages as read on 13 October 2026: the developer quickstart, the API reference, production best practices, the rate limits, spend limits and error codes guides, the pricing page and the help centre's prepaid billing article. Prices are US dollars. Some other help centre articles refused automated reading that day, so claims that rested on them have been removed.

What you need first

• An OpenAI account. The sign-in page accepts an email address, Google, Apple, Microsoft or a phone number.
• A payment method for prepaid credits. A ChatGPT plan does not help here: OpenAI's plan docs say API token prices are separate from subscription usage.
• A supported country or territory. Requests from elsewhere return a 403 error.

New to the term? An API is how one program asks another to do something; the key is the password that says who is asking. See API in the glossary.

How to get an OpenAI API key, step by step

Step 1: Create the key

Sign in and open the API keys page. Create a new secret key and name it after where it will run, such as n8n-production. Copy it and store it somewhere safe straight away. OpenAI's best practices guide strongly recommends setting an expiration date when you create a project key and keeping a regular rotation routine.

If you cannot create one, an administrator may have restricted it. Organizations can allow only service-account keys, allow only user-owned project keys, or disable new key creation, and can enforce a maximum key lifetime.

Step 2: Add prepaid credits

New API accounts use prepaid billing. From OpenAI's prepaid billing article:

1. Open the billing overview for the organization you want to fund.
2. Select Add payment details and complete the payment step.
3. Choose the first purchase. The minimum is $5 and the default is $10.
4. Review Use auto-reload, which is turned on by default. Turn it off, or set a threshold and a monthly reload limit.
5. Confirm. The balance can take a few minutes to update.

Purchased credits expire after one year and are non-refundable outside a short list of exceptions, so start small.

Step 3: Save the key as an environment variable

The quickstart exports the key so that each SDK reads it from the system environment:

# macOS / Linux
export OPENAI_API_KEY="your_api_key_here"

# Windows (PowerShell)
setx OPENAI_API_KEY "your_api_key_here"

On Windows, open a new terminal after setx so the variable is visible. On macOS or Linux, add the export line to your shell profile to keep it between sessions.

Step 4: Test the key with one request

Install the SDK with pip install openai or npm install openai, then run the quickstart example. In Python:

from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-6-astra",
    input="Write a one-sentence bedtime story about a unicorn.",
)

print(response.output_text)

Or in JavaScript, saved as example.mjs and run with node example.mjs:

import OpenAI from "openai";
const client = new OpenAI();

const response = await client.responses.create({
  model: "gpt-6-astra",
  input: "Write a one-sentence bedtime story about a unicorn.",
});

console.log(response.output_text);

A reply means the key, billing and model access all work. Both examples are copied from the quickstart, which uses gpt-6-astra, the most expensive of the three featured models. For a test, swap in gpt-6-luna. If you see a 401 or a 429 instead, go to troubleshooting.

Which kind of credential you need

OpenAI's API reference lists three ways to authenticate, and its best practices guide adds service accounts. Most readers need the first row.

If you belong to more than one organization, or use a legacy user key, pass the OpenAI-Organization and OpenAI-Project headers so usage is billed to the right place.

What the OpenAI API costs

You pay per token. Standard rates per million tokens for the three models OpenAI features on its models page, from the pricing page:

Prompts with more than 272K input tokens are priced at twice the input rate and 1.5 times the output rate for the whole request. Batch and Flex processing cost half the standard rate when you can wait for results. For scale, a reply of 1,000 output tokens costs $0.0005 on GPT-6 Luna and $0.05 on GPT-6 Astra. To price your own prompt per request and per month, use the free OpenAI API cost calculator, which also lists Claude and Gemini models. Our LLM API pricing comparison sets these beside other vendors.

Usage tiers

Your organization's tier rises automatically as its total credit purchases grow, and higher tiers bring higher rate limits. OpenAI cut the paid tiers from five to three on 6 October 2026.

The usage limit is a ceiling OpenAI assigns, not credit it gives you. Rate limits apply per organization and per project, not per user, and vary by model: on the Build tier GPT-6 Luna allows 5,000 requests and 2,000,000 tokens a minute. Your own numbers are under Organization limits.

Set a hard spend limit before you ship

A spend alert only notifies you. A hard spend limit makes requests fail with a 429 once tracked spend reaches your cap, which bounds the cost of a runaway loop or a leaked key. From the spend limits guide:

1. Go to Organization limits, or Limits in a project's settings.
2. In Spend, select Edit spend limit and enter the monthly amount.
3. Turn on Enforce a hard limit and select Save.

Security checklist

Troubleshooting common errors

The common failures are all in OpenAI's error codes guide. For anything billing-related, read error.code: the broader type can still say insufficient_quota.

For rate limits, wait at least as long as the Retry-After header says. If it is missing, use exponential backoff with jitter and cap the number of retries. Building retries into a workflow? Our n8n error handling guide shows the pattern.

Using your key in n8n

In n8n the key lives in a reusable OpenAI credential, not in each node. Per the n8n docs, paste it into API Key and leave Organization ID blank unless you belong to several organizations. The same credential works for the OpenAI, Chat OpenAI and Embeddings OpenAI nodes. Give each client or workflow its own key, ideally in its own project, so you can cap or revoke one without affecting the others.

From there, connecting OpenAI to n8n covers full workflows, ChatGPT API workflows covers calling the API directly, and the free n8n cron expression generator helps you schedule them.

Where to go next

Many projects hold a key from each vendor and send each task to the model that suits it. The Anthropic side is in how to get a Claude API key, and Claude vs ChatGPT and Claude vs GPT compare the two. If you plan to build a product on a model API, AI SaaS Builder, one of our programs, does it on the Claude API with Supabase, Vercel and Stripe. The AI automation library has the rest of our guides.

What changes often

Model IDs, prices and tier names all moved in the weeks before this was written: GPT-6.1 Sol arrived on 29 September 2026 and the usage tiers were cut from five to three on 6 October. Dashboard labels change too. The steps and the security advice are stable. Check the linked page before you rely on any number here.

OpenAI API key: FAQ