Pricing & tokens
Cloud measurements are charged in tokens — input per second plus a fixed price per enabled metric set. Test keys and on-device are free.
Cloud measurements are billed in tokens from your organization's balance. The reference rate is $1 = 100,000 tokens. Only cloud measurements are charged — test keys, the on-device tier and every other API call are free.
The live configuration is always available at the public
GET /api/v1/pricing; the numbers below are the current reference values.
The same numbers, with token packs, live on the public
pricing page.
The formula
A completed cloud job charges input + output:
- Input —
1,000 tokens × ceil(seconds processed). Heart rate, breathing rate, stress and signal quality are always included. - Output — a fixed price per metric set the job requested in
features:
| Set | Tokens per measurement |
|---|---|
hrv — SDNN, RMSSD, pNN50, frequency analysis | 5,000 |
bp — blood pressure estimate (beta) | 5,000 |
spo2 — SpO2 estimate (beta) | 3,000 |
emotion — emotion recognition (beta) | 2,000 |
gaze — gaze and head pose (beta) | 2,000 |
A full 30-second measurement with every set enabled costs 30,000 + 17,000 = 47,000 tokens ≈ $0.47.
When the charge happens
The charge lands when the job settles as completed — in the same transaction
that credits your cloud seconds. failed, aborted and expired jobs are
free. When a dropped stream falls back to batch, only the batch job that
actually produced the measurement is charged.
The balance check at minting is an estimate (for batch jobs, 60 seconds of input plus the requested sets). A running measurement is never cut off: the final charge may take the balance slightly below zero, and only the next minting is blocked.
Errors
| HTTP | Code | Meaning |
|---|---|---|
| 403 | token_quota_exceeded | The key's monthly token cap (tokenQuotaMonth) is exhausted |
| 403 | insufficient_tokens | The organization balance cannot cover the estimated charge |
Set a per-key cap in Console → Keys → Limits & scopes to keep one integration from spending the whole balance. The console asks for it in dollars and converts at the current token rate.
Free allowance and top-ups
- Every new organization starts with a welcome grant of 1,500,000 tokens — about 30 full measurements.
- Keys created in test mode return deterministic synthetic results and never spend tokens.
- Spent tokens are not refundable; an unused balance can be refunded within twelve months of the payment. See the refund policy.
Paying by card (Paddle)
Token packs are bought in Console → Spend. The checkout is handled by
Paddle as the merchant of record — it charges the card, issues the invoice
and handles VAT. The balance is credited when Paddle's webhook reports the
transaction as paid, so the tokens appear once that notification arrives — we
do not control how long Paddle takes. The credit lands in the same ledger as
every other movement, with the reason purchase. For custom volumes write to
us — a manual credit uses the same ledger.
Next: Limits & errors.