How to get an OpenAI API key (the ChatGPT API key)

Get an OpenAI API key in 6 steps

1. Sign in or create an account at platform.openai.com.
2. Open the API keys page at platform.openai.com/settings/organization/api-keys and create a new secret key.
3. Copy the key right away, because OpenAI shows the full secret only once when you create it, and if you lose it you'll have to create a new key and update your application.
4. Go to your organization's billing overview, add your payment details and buy at least $5 of prepaid credit (the default is $10); auto-reload is turned on by default during setup, so switch it off if you don't want automatic purchases.
5. Save the key in the OPENAI_API_KEY environment variable by adding export OPENAI_API_KEY='yourkey' to ~/.zshrc or ~/.bash_profile on macOS and Linux, or by running setx OPENAI_API_KEY "<yourkey>" in the Windows command prompt.
6. Send a test request to make sure the key works.

What it costs to use the key

Creating an OpenAI API key is free, but API usage is billed separately from any ChatGPT plan.

• Monthly usage limits cap what you can spend by tier: $100 on Free, $500 on Build (after $5 in total credit purchases), $5,000 on Launch (after $100) and $200,000 on Grow (after $500).
• Your balance can briefly go negative because API access may not stop the instant your credits run out, and that negative amount gets deducted from your next purchase.

The table below shows which accounts and ChatGPT plans include API credits: only paid Pro and Business plans do.

Make your first request

Every request sends your key in an Authorization: Bearer header, and the official SDKs (npm install openai or pip install openai) read OPENAI_API_KEY from your environment, so you never hardcode it. Revoking a key takes effect within a few seconds, while most other changes to a key's authentication propagate within 15 minutes.

The first call below checks that your key works by listing models at https://api.openai.com/v1/models, and the second sends a prompt to the Responses API (POST https://api.openai.com/v1/responses).

export OPENAI_API_KEY="your-key-here"

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

curl https://api.openai.com/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{"model": "gpt-6.1-sol", "input": "Say hello in five words."}'

Keep the key safe

• Use one key per person: Sharing API keys is against OpenAI's Terms of Use, so invite team members to your organization instead.
• Never ship a key in a browser or mobile app: Route all API requests through your own backend.
• Never commit a key to a repository: Keep it in an environment variable or a key management service.
• Set an expiration date on project keys and rotate them: Admins can enforce a maximum key lifetime for the organization or a project.
• Set monthly spend limits with hard enforcement: Spend alerts alone won't stop API traffic when your budget runs out.
• Rotate immediately if a key may have leaked: You can also turn on IP allowlisting so OpenAI rejects requests from other addresses even when the key is valid.

Fix common API key errors

The table below lists the authentication, billing and limit errors from OpenAI's error codes guide. Only the 401 errors point at the key or your IP, while four of the 429 errors come from billing and spend limits. For billing errors, check error.code in the response, because error.type can still read insufficient_quota.

Use your ChatGPT plan as the API key

If what you want is a ChatGPT API key that runs on the ChatGPT subscription you already pay for, Plan2API does that for ChatGPT plans with Codex, such as Plus or Pro, without per-token billing.

You sign up with a Google account, connect ChatGPT by signing in with a one-time device code, and get a p2a_ key that works with the OpenAI Chat Completions format (POST /v1/chat/completions) at https://api.plan2api.com/v1 using model ids like openai/gpt-6.1-sol, openai/gpt-6-astra or openai/gpt-6-luna. Requests run on your plan's own Codex limits, there's no /v1/responses or /v1/embeddings endpoint, and the Plan2API Free plan gives you 50 chat requests a day for $0 with no card required.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.plan2api.com/v1",
    api_key="p2a_...",
)

reply = client.chat.completions.create(
    model="openai/gpt-6.1-sol",
    messages=[{"role": "user", "content": "Write a haiku about API keys."}],
)
print(reply.choices[0].message.content)