Frequently asked questions

LuxeChain speaks both the OpenAI and the Anthropic format, so most tools work by changing one setting. Authenticate with your API key (lxc_live_…) as a Bearer token.

• OpenAI SDK, and tools that use it (Cursor, n8n, and similar) — set the base URL to https://api.luxechain.si/v1
• Anthropic SDK and Claude Code — set the base URL to https://api.luxechain.si (no /v1 on the end)

Call GET /v1/models with your key to see the models that are live right now. The full endpoint reference and code samples are in our documentation.

The live catalogue is on the Models page, and GET /v1/models returns the same list as JSON. We add models as providers release them. Nothing is locked to a single vendor, and you do not need to re-provision anything to switch — the same key reaches all of them.

OpenAI-compatible: https://api.luxechain.si/v1
Anthropic-compatible: https://api.luxechain.si

Your key is the same for both. Copy the exact string — a trailing slash or a missing /v1 is the most common cause of a connection error.

Usage is deducted for the actual work the model reports, or per generation for image, video and audio. Each plan is a monthly pool of usage credits, and every request is deducted from that one balance, whichever model or modality you use. Heavier models spend more credits for the same amount of text, and the pricing page shows what each model buys. Failed or rejected requests are never charged. Credits are valid for 30 days from the day they are credited.

Yes. Your account page has a live activity feed that refreshes every few seconds: every request, the credits it used, what it cost, and your remaining balance. Nothing is batched or delayed until the end of the month.

No. You are charged the exact cost of the request. Very small requests cost very little; nothing is rounded up to a floor.

No. Credits are valid for 30 days from the day they are credited, and they are spent soonest-expiring first, so nothing lapses while fresher credits sit idle. Pick the plan size that matches what you actually use — see the Refund & Cancellation Policy for how expiry interacts with refunds.

Requests stop with an out-of-credits response until you top up. Move up a plan from your account page and the extra credits are credited as soon as the payment clears.

• Rate limit: 60 requests per minute per key, up to 180 requests per minute per account.
• Concurrency: 5 simultaneous requests per account.
• Payload limit: up to 25 MB per request.

These exist so one heavy user cannot slow the Service for everyone. If you have a workload that needs more, talk to us — we would rather plan it with you than have you hit a wall.

Requests are streamed straight through as the model produces them, so time-to-first-token is what you would see calling the model directly. We do not buffer responses.

No. LuxeChain operates strictly as a gateway: prompt contents and model outputs are streamed through directly and are never saved on our servers. Only usage metadata — the model, token counts, cost and timestamp — is recorded, because that is how your billing and your activity feed work. Full detail is in our Privacy Policy.

No. We do not train models, and we do not sell or rent your personal data.

Your account details, usage metadata, and the payment records the law requires. Retention periods are set out in the Privacy Policy.

Create an account, choose a plan, and your key is issued when the payment clears. You can rotate it from your account page at any time — rotating invalidates the old key immediately.

Rotate it from your account page straight away, then tell us through our contact form. Rotating stops the old key working, so a leaked key cannot be used against your balance.

Any time, from your account page or by email. No further payments are taken, and you keep access until the end of the period you have paid for. See the Refund & Cancellation Policy.

Use our contact form. Include the model, the time, and the error text — with those we can find the request in the metering log quickly. Check the status page first in case it is a known outage.