Common questions

Billing & credits

Is this a subscription? No — it's prepaid. You top up a USD balance and usage draws it down at each model's per-token price. There's no recurring AI charge beyond what you use.

What happens when the balance runs out? Inference requests are blocked with an insufficient credits error until you top up. Nothing is lost — keys, agents, settings, and history all remain — and requests resume within moments of a top-up. Enable auto top-up to avoid interruptions.

Can I get a refund of unused balance? Raise a support request. Refunds appear in the transaction history as their own type.

Do I get receipts? Yes — every top-up has a receipt number and a downloadable PDF in the transaction's details drawer, and the receipt is emailed to your facility's billing contacts.

Why does "Spent today" differ from the transaction history? Usage charges are reconciled into the ledger every 15 minutes, while the spend figures and available balance are live. They converge on the next reconciliation.

How is the Auto Router priced? It has no price of its own — you pay the rate of the model it picks for each request. The logs show which model that was.

Privacy & data

Are my prompts or the AI's responses stored? No. Activity and Logs record metadata only — model, tokens, cost, latency, status, caller. Prompt and response content is never stored by Performance Hub.

Can I stop personal data reaching the models? Yes — enable PII masking under Settings → Guardrails, and add custom content filters for anything domain-specific. See Guardrails.

Can I restrict which providers my data goes to? Yes — the Models page has one-click presets (e.g. Western providers, described as "data-sovereignty friendly") and per-model toggles. Keys and agents can be restricted further. See Models.

API keys & integrations

What software can I connect? Anything that speaks the OpenAI API and lets you set a custom base URL and API key (BYOK) — the OpenAI SDKs, most AI frameworks, and many off-the-shelf tools. See Connecting your application.

Can I see the key secret again after creating it? No — it's shown exactly once. If it's lost, Rotate the key for a new secret.

How many keys should I create? One per system. Labels make each system's usage visible in Activity and Logs, and a compromised key can be rotated without touching anything else.

What's the difference between revoking and deleting a key? Revoke disables the key immediately but reversibly (Re-activate restores it). Delete is permanent.

Can one key be limited to specific models? Yes — Restrict models for this key in the key's configuration. The same applies per agent.

AI Agents

Do my AI Agents need an API key? No. Agents paired with your facility connect automatically with per-device credentials that Performance Hub manages. They appear under Settings → Agent Access & API Keys with their own spend and controls.

Can I cap what a single agent spends? Yes — per-agent daily, monthly, and lifetime limits, plus rate limits, model restrictions, and guardrail overrides. See AI Agents.

Models & routing

Which model should my integration use? auto (the Auto Router) is the recommended default — it routes each request to a complexity-appropriate model, so simple work goes to fast, cheap models and hard tasks get the strongest ones. Or use any specific model ID enabled for your facility.

Why do my logs show a different model than I requested? You requested auto — the logs show the model the router resolved to, and the request drawer records auto as Requested model. Expected behaviour.

Can I change what auto routes to? Yes — Edit routing on the Auto Router card lets you pick the model for each complexity tier for your facility, with Reset to default available. See Models.

A model I want isn't in the catalog. Use the Raise a support request link on the Models page: "Want to see a model added? Raise a support request and we'll add it if it's generally available."

Limits

What's the difference between the balance and a spend limit? The balance is money you've credited; spend limits are policy caps you choose. Either can block requests: an empty balance blocks until you top up; a hit limit blocks until its window resets (or you change it). See How it works.

When exactly do daily and monthly limits reset? Daily at midnight and monthly on the 1st — both in the facility's local timezone. The budget exceeded error message states the precise reset time.

Does the lifetime hard cap ever reset itself? No. It counts spend since its last manual reset. Reset hard cap starts the counter over from $0 (the cap amount is unchanged) and saves the closed period to the reset history.

Do per-key limits and account limits stack? Yes — both apply. A request must pass the account limits and the calling key's/agent's own limits (and the balance) to run.