Pay for answers, not attempts

Every enrichment that returns a merchant counts once, whether it came from the cache or the AI. Results with merchant: null are free, though they still count toward your rate limit. In a batch, each item that returns a merchant counts once.

Enrichment requests return HTTP 402 with the code quota_exceeded until the next month starts (quotas reset on the 1st, UTC). A batch larger than your remaining quota is rejected as a whole, so you never get half a batch.

Each API key has a requests-per-second limit set by its plan, with a small burst allowance. Above it the API returns HTTP 429 with a Retry-After header telling you when to try again.

We only need the descriptor, never card numbers or cardholder names. Amount and currency are validated but not stored. Your recent requests are kept so you can debug them in your dashboard, and recognised descriptors are cached to answer faster. Person-to-person transfers are never cached or kept. Descriptors we could not recognise are kept for up to 30 days, without your account details, so we can improve recognition. When the AI fallback runs, it receives only the cleaned merchant text with provider-side storage turned off, and we do not train models on your data.

All of them. Descriptors are read the same way whichever bank or country they come from: we recognise 252 countries by code and name, 1,335 cities with their airport codes and local spellings, regional codes such as CA, ON or NSW, and Arabic-script descriptors.

No. Sign up with an email address and you get an API key with 500 enrichments a month.