Bringing your own API keys
If you're on the BYOK plan, the AI in Studio runs on your own Anthropic and OpenAI API keys. You pay those providers directly for usage, at their rates, on their invoices. Studio only charges the flat $20/mo subscription on top.
This page covers setting up and managing those keys. If you're still on the Managed plan and want to switch, see Plans.
Where keys live
Open Settings and look for the API Keys section (API Key Management). It has two provider panels:
- Anthropic — Claude AI models (Haiku, Sonnet, Opus).
- OpenAI — GPT models (GPT-5.4 and its smaller variants).
You can configure either, or both. With both set, the AI will use whichever matches the model you've selected in the chat.
If you're not on the BYOK plan, the section is visible but reminds you:
API keys are used with the BYOK plan. Upgrade to BYOK above to use your own keys.
Getting your keys
Studio doesn't sell or proxy keys — you create them in your own provider accounts:
- Anthropic: sign in at
console.anthropic.com → API Keys
→ Create Key. Anthropic keys start with
sk-ant-…. - OpenAI: sign in at
platform.openai.com →
Create new secret key. OpenAI keys start with
sk-…(newer ones beginsk-proj-…).
You'll need a payment method set up with each provider for the keys to work — Studio doesn't pre-pay them on your behalf.
Adding a key
In Studio's API Keys section, paste the key into the provider's input
and select Save Key. Studio masks the key after saving — you'll see
something like sk-ant-api03-*** with a Connected badge.
Testing the connection
After saving (or any time later), select Test Key. Studio makes a tiny real call against the provider:
- If it works, you'll see Key verified successfully.
- If the key is rejected, you'll see Invalid API key. Please check the key and try again. Common causes: a typo, a key you've since revoked, or no payment method on file with the provider.
How keys are stored
Keys are stored encrypted on Studio's servers using AES-256-GCM. Only the first few characters are shown back to you (everything else is masked). Nobody at VectrForce reads the plaintext; the AI service decrypts it just long enough to make the request to your provider.
What if I only configure one provider?
You can. If you set only an Anthropic key, you'll be able to use Claude models but not GPT models. If you set only an OpenAI key, you'll be able to use GPT models but not Claude. You can always come back later and add the other.
If the model you pick in the chat isn't covered by any of your keys, the AI won't be able to run. The chat will show an error — either add the missing key or switch to a model you have a key for.
DeepSeek and BYOK
DeepSeek isn't available on BYOK. It runs only on Studio's platform key. If you want to use DeepSeek's low-cost option, stay on the Managed plan (or switch back to it for that work).
Updating a key
Paste a new key into the same field and select Save Key again — the old one is replaced. You'd typically do this if you rotated the key in your provider account.
Removing a key
Select the trash icon next to a saved key, then Confirm. Removal is immediate. If a chat is mid-run when you remove the key it's using, the AI will fail on its next request — give it a moment, then ask again with a key in place.
Per-project keys
Studio supports per-project keys at the API level (so a project can use a different key than your account default), but the current settings UI only exposes account-level keys. For most users, one set of keys at the account level is what you want.
Next
- Plans — switching between Managed, BYOK, and Dedicated.
- Choosing an AI model — picking among the models your keys cover.
- How AI usage is billed — billing details for both plans.