Get an OpenAI API key
Step-by-step: create an OpenAI Platform account, add a payment method, mint an API key, and drop it into ShellYard Settings → Magellan → OpenAI.
ShellYard’s Magellan AI requires your own API key when you select the OpenAI provider. Keys are minted in the OpenAI Platform dashboard — separate from ChatGPT consumer accounts, even if you already pay for ChatGPT Plus.
Steps
-
Create an OpenAI Platform account. Go to platform.openai.com and sign up with email, Google, Microsoft, or Apple. ChatGPT Plus does not include API credits — Platform is a separate billing surface from the chat product.
-
Add a payment method. API usage is pay-as-you-go. Navigate to Settings → Billing → Payment methods and add a card. Setting a monthly hard limit under Settings → Limits is recommended for first-time API users — it caps spend so a runaway script can’t surprise you.
-
Mint an API key. Open Dashboard → API keys → + Create new secret key. If your account has Projects enabled, scope the key to a project so you can revoke it independently of other apps. Name it something you’ll recognize — e.g.
shellyard-laptop. The key is shown once; copy it immediately. -
Paste it into ShellYard. Open Settings → Magellan → Provider: OpenAI → API key and paste. ShellYard fires a small test request to confirm the key works, then stores it (see BYO key for storage details by tier).
Key format
Modern keys are project-scoped and start with sk-proj-. Legacy account-wide keys starting with sk- (no -proj- segment) remain valid but aren’t issued to new accounts. Both work with ShellYard.
Picking a model
ShellYard exposes a smart model + fast model pair under the provider settings — Magellan routes per stage. Typical pairings:
| Tier | Smart model | Fast model |
|---|---|---|
| Best quality | GPT-5 | GPT-5 mini |
| Balanced (default) | GPT-5 mini | GPT-5 nano |
| Lowest cost | GPT-5 nano | GPT-5 nano |
The exact model identifiers update over time. See platform.openai.com/docs/models for the current list and context-window sizes.
Pricing
OpenAI prices per-token, with separate input, cached-input, and output rates. The current rate card is at openai.com/api/pricing. The cached-input pricing is meaningful for Magellan because the five-stage flow keeps a stable context across turns — repeated prefixes are billed at the cached-input rate, which is typically 50% of the standard input rate.
A monthly hard limit in Settings → Limits is the safest way to keep spend bounded.
Rotating
If a key leaks, return to Dashboard → API keys, revoke the compromised key, mint a new one, and overwrite the value in ShellYard. The revoked key stops working within seconds — no need to coordinate.
Troubleshooting
- “401 Unauthorized” on first use — usually means the key was copied with a trailing space, OR the billing setup hasn’t fully propagated. Try minting a fresh key after confirming a payment method is attached.
- “429 Too Many Requests” — you’ve hit your monthly spend limit. Increase the limit under Settings → Limits or wait for the limit window to reset.
- Key works in
curlbut not in ShellYard — check Settings → Magellan → Provider is set to OpenAI, not OpenAI-compatible. The compatible provider takes both a key and a base URL pointing at your self-hosted endpoint.