How to Get an OpenAI API Key in 2026: Dashboard, Billing, Environment Variable and First Call

Quick answer: Sign in at platform.openai.com, open the API keys page (platform.openai.com/api-keys) and create a secret key. Save it as OPENAI_API_KEY, install the official SDK and send one request to the Responses API. OpenAI lets you run a free test request; after that, add credits on the billing page. Per-token rates are on OpenAI’s pricing page (accessed October 25, 2026).

What do you need before you start?

An account on the OpenAI platform (platform.openai.com). API billing is separate from ChatGPT: a ChatGPT Plus or Pro subscription does not include API usage. See OpenAI API pricing for how the model tiers compare.

How do you create an OpenAI API key, step by step?

1. Sign in at platform.openai.com.
2. Open the API keys page (platform.openai.com/api-keys) and create a new secret key. Pick the project it should belong to; usage is tracked per project.
3. Copy the key right away and keep it somewhere safe, such as a password manager or secret store.
4. Export it as an environment variable, as OpenAI’s quickstart shows:

# macOS / Linux
export OPENAI_API_KEY="your_api_key_here"

# Windows PowerShell
setx OPENAI_API_KEY "your_api_key_here"

5. Make a first call with the official Python SDK (pip install openai):

from openai import OpenAI

client = OpenAI()  # reads OPENAI_API_KEY from the environment

response = client.responses.create(
    model="gpt-6-astra",
    input="Say hi in five words.",
)
print(response.output_text)

The quickstart uses gpt-6-astra. For a cheap smoke test you can swap in a lower tier from the pricing page. Node.js works the same way with npm install openai.

6. Add credits on the billing page (platform.openai.com/account/billing/overview) once the free test request is done.

How do you call the API without an SDK?

Send the key as a Bearer token. OpenAI’s API reference shows:

curl https://api.openai.com/v1/models \
  -H "Authorization: Bearer $OPENAI_API_KEY"

If you belong to more than one organization, or use a legacy user key, add the OpenAI-Organization and OpenAI-Project headers so usage lands in the right place. You can find both IDs in your dashboard settings.

Which kind of key do you need?

OpenAI’s reference lists three ways to authenticate:

Credential	Use it for
Standard API key	Normal application requests (most people)
Admin API key	Administration endpoints: users, invites, projects, API keys, audit logs
Workload identity federation	Short-lived access tokens for cloud workloads, instead of a long-lived key

Can you get an OpenAI API key for free?

The key costs nothing, and the quickstart includes a free test request. Real use is pay as you go, billed per token. Avoid “free OpenAI API keys” posted online: a key is tied to its owner’s billing account, leaked keys get revoked, and using someone else’s key can break OpenAI’s terms.

What if the key doesn’t work?

• 401 “Incorrect API key provided”: re-copy the key, and check that an old OPENAI_API_KEY in your shell or system environment is not overriding your .env file.
• 429 or insufficient_quota: you are out of credits or over your rate limit. Add credits or slow down.
• Key was just revoked or changed: revocations apply within seconds; other auth changes can take up to about 15 minutes to propagate.

Full fixes for every status code are in how to fix OpenAI and Claude API errors.

How do you keep the key safe?

OpenAI’s rule: your key is a secret, so “don’t share it with others or expose it in any client-side code such as browsers or apps.” Load it on the server from an environment variable or a key management service, never commit it to Git, and revoke and replace it if it ever leaks.

Using other providers too? See how to get a Claude API key and how to get a Gemini API key.