How to get an OpenAI API key: sign up, billing, first curl, and rate limits

Register on the OpenAI platform, set up billing, create an API key, send one curl request, and understand 429 rate-limit errors.

You want GPT in code, not only in the ChatGPT web app. The official path is platform.openai.com: sign up → set up Billing → create an API key → send your first curl to chat/completions. Steps follow the OpenAI Quickstart and Authentication docs.

The searchable job is how to get an OpenAI API key and run your first API request: account, billing, key, a curl example, and how to read rate limits. Trust Billing / Limits in your account over third-party quota tables.

Platform vs ChatGPT

Sign up through first curl

1. 01
   Sign up at platform.openai.com

   Open platform.openai.com, register with email or supported OAuth, verify email, and land in the Dashboard. Quickstart lists Sign up / Log in as step one.

2. 02
   Set up Billing

   Go to Settings → Billing and add a supported card or payment method. API usage is metered by tokens; see Pricing. Calls may fail until Billing is ready.

3. 03
   Create an API key

   Settings → API keys → Create new secret key. Store it in a password manager. Authentication docs treat keys like passwords—never commit to Git or ship in frontend bundles. Create separate keys per project and revoke individually.

4. 04
   Send your first Chat Completions curl

   Export $OPENAI_API_KEY, then POST to https://api.openai.com/v1/chat/completions. On success, choices[0].message.content holds the model reply.

5. 05
   Read rate limits and 429

   The Rate limits guide explains RPM and TPM caps per model. Response headers include x-ratelimit-remaining-requests. On 429, honor retry-after or lower concurrency; long-term caps live under Settings → Limits.

export OPENAI_API_KEY="sk-..."  # never commit

curl https://api.openai.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "Say hello in one word."}]
  }'

Where first calls usually break

401 Unauthorized

Missing, mistyped, or revoked key. Confirm Authorization: Bearer and Active status under Settings → API keys.

429 Too Many Requests

Hit RPM/TPM caps. Read headers and retry-after; Rate limits docs recommend exponential backoff.

403 or model not found

Your usage tier may not include that model. Pick a default from Quickstart or check Limits.

Run one minimal curl before wiring Cline, Cursor, or your own service. A leaked key costs more than picking the wrong model—rotate instead of chasing invoices.