Answers are generated on your account
Replies are sent to your provider instead of our AI, and the usage appears on your provider's bill. If you set a default model on the key, that model is used.
AI
No, you do not need your own AI key — it only lifts your limit.
Every plan answers on our AI from day one, and no plan requires a key of yours. Adding one moves model usage to your own provider account and lifts your monthly customer limit. And the phrase "bring your own keys" means two unrelated things in this product. Connecting WhatsApp does not give the AI a key. That confusion is the single most expensive one here, so it is the first thing on this page.
Read this first
Two different things are both called “bring your own keys”
They live on different screens, they do different jobs, and one never stands in for the other. A business that connected WhatsApp, sawConnected, and concluded “keys are done” has done half of one of these and none of the other.
| Channel credentials | AI provider key | |
|---|---|---|
| What it is | Your own WhatsApp Business account, your own Telegram bot, your own Facebook or Instagram app. | An API key from an AI company: OpenAI, Anthropic (Claude), Google Gemini, or anything that speaks the OpenAI API. |
| What it buys you | Messages arrive and go out under your number and your business name. | The words the assistant writes are generated on your AI account, billed by that provider. |
| Where you add it | Dashboard → Channels (in the Setup group of the sidebar). | Dashboard → Provider keys (in the Account group of the sidebar). |
| Who needs it | Anyone who wants WhatsApp, Telegram, Instagram or Messenger. Optional if you only use the website widget. | Never required. Optional on every plan, and on a paid plan it lifts the monthly customer limit. |
The trap, stated plainly: connecting WhatsApp, Telegram or Instagram does not give the AI a key. Only a key on the Provider keys screen counts. Nothing on the Channels screen is ever read as an AI key.
The short answer
Do I need one to start?
- Free and Standard: no. The assistant answers on our AI as soon as your domain is verified and the widget is in. You can add your own key later at any time.
- Why add one at all? Two reasons: a premium model of your choosing, and — on a paid plan — usage that runs through your provider account, which lifts the monthly customer limit.
Adding a key changes where model usage is billed, never whether the assistant answers. If your key stops working — revoked, out of credit, rate-limited — we fall back to answering on our own AI rather than leaving your visitors without a reply. You are told when that happens.
Being honest about today's behaviour: the hard stop at the end of that window is currently switched off at the platform level. We would rather cover a small amount of AI cost than have a live business's chat go quiet. Do not build on that — it is a deliberate temporary posture, not a promise, and the plan still expects your key.
What changes
What actually happens when you add a key
If your provider fails, we quietly carry on
A rejected key, a spend cap, an outage — we fall back to our own AI so the visitor still gets an answer. That is good for your customers and confusing for you: a broken key looks like “the answers changed”, not like an error. The Last used line on the Provider keys screen is how you check whether your key is really being used.
Your plan's daily allowance still applies
This surprises people, so it is worth being blunt: adding your own key does not switch off the daily token counter on your plan. Usage is still counted, and while you are between 80% and 100% of the day's allowance, replies are forced onto a lighter, cheaper model even when your key is set. See the daily limit page.
One key per provider, and a fixed order of preference
You can store a key for each provider and switch each one Active or Off. If more than one is Active, the order used is OpenAI, then Anthropic, then Gemini, then OpenAI-compatible. To force a particular provider, leave only that one Active.
The key itself is not shown again
Keys are encrypted before storage. After you save, the screen shows only a short preview, so keep your own copy wherever you keep passwords.
How
Add a key
- Sign in to your dashboard at
app.orcalinq.com. In the sidebar, under Account, open Provider keys. Only the account owner sees this item — see below if it is missing. - Under Add a provider key, choose the Provider: OpenAI, Anthropic (Claude), Google Gemini, or OpenAI-compatible (self-hosted / other).
- Paste the key into API key.
- Optional: a Default model. Leave it empty and the provider's own default is used.
- For OpenAI-compatible only, a Base URL is required. The form will refuse to save without one.
- Press Test connection if you want to check it first. A pass reads “Key works — the provider accepted it.” A fail shows the provider's own reason.
- Leave the tick-box Verify the key with the provider before saving ticked, then press Save key. That tick-box — not the Test button — is what earns the verified badge in the list.
Answers route through the key from the next visitor message. There is nothing to restart, and the widget does not need reinstalling.
Where to get the key
Inside the AI provider's own console
These are other companies' screens and they get rearranged without notice. If a label below is not where we say, use the search box in that console for the words in bold — the underlying page rarely moves, only the menu around it.
OpenAI
Sign in at platform.openai.com (the developer platform, not the ChatGPT app) → API keys → Create new secret key. Copy it immediately; it is shown once. A key starting sk- is the right kind.
No API keys item? You are probably in the ChatGPT consumer site. A ChatGPT Plus subscription is not an API key and will not work here — the API is billed separately and needs credit on the account.
Anthropic (Claude)
Sign in at console.anthropic.com → API keys → Create key. Keys start sk-ant-.
A Claude.ai chat subscription is a different product from the Console and carries no API key.
Google Gemini
Sign in at aistudio.google.com → Get API key → Create API key, and pick a Google Cloud project when asked.
If you are sent into the Google Cloud console instead, the key you want is under APIs & Services → Credentials, with the Generative Language API enabled on that project.
Something else
Any service that copies the OpenAI API — a self-hosted model, a regional provider — goes in as OpenAI-compatible with its base URL. If you are unsure what the base URL is, it is the address their documentation puts before /v1/chat/completions.
If that did not work
The five things that actually go wrong
- No “Provider keys” in the sidebar. It is owner-only, deliberately, because the key bills the owner's own provider account. A teammate with dashboard access cannot see or add one. Ask the account owner, or sign in as the owner.
- Test connection fails. The message is the provider's own words, and the usual causes are a key copied with a trailing space, a key from the wrong account, or no billing credit on the provider. Fix it there, then test again — do not save a key you know is rejected.
- “Key saved (not verified)”. The key is stored and will be used; only the check did not run or did not pass. Paste it again with Verify the key with the provider before saving ticked if you want the verified badge.
- Nothing about the answers changed. Check Last used on the key. If it stays empty, we are falling back to our AI on every call — usually a rejected key or an exhausted provider balance. Also check the key is switched Active and not Off.
- A warning bar about your key will not clear. The bar clears once a key is Active. If it persists, it is naming a different problem — read the bar's own sentence, which always says both the cause and the fix.
Still stuck? Write to support@orcalinq.comwith your provider name and the exact wording of the error. Never send us the key itself.
Start on our AI. Move to your own key when it suits you.
Nothing about the setup changes when you switch — same widget, same channels, same knowledge base.