Most web-data tools charge a flat fee per request, so a cheap static fetch quietly subsidizes an expensive AI extraction — or you overpay for the simple calls. Synoppy prices every call by the work it actually does. Here's how the meter runs.
Two ways a call is metered
Operations split into two buckets, because their real costs are nothing alike:
- Static reads — Read, Crawl, Map, Enrich, Images — are volume-metered, at roughly one credit per 150 KB of content processed. A short page is a single credit; a long one scales up gently, so you pay for the bytes you actually pull.
- Compute-heavy ops — Extract, Classify, JS rendering, and Screenshot — are cost-metered, priced from the real tokens an AI call consumes or the real browser-time a render takes. A small extraction costs a few credits; a heavy one costs proportionally more.
Everything draws from one pooled balance, so you never pre-commit budget to a product you don't use this month. And you never have to guess: every response reports creditsUsed and creditsRemaining, so the meter is visible on every call.
Why metered beats flat
A flat per-endpoint price has to assume the worst case, so it overcharges the common one. Metering lets a 2 KB API doc cost a single credit while a 400 KB article costs a handful — and an AI extraction is billed on the tokens it truly spent, not a round number someone picked.
Refills
Paid plans refill to their monthly allotment on each billing cycle — credits don't roll over or stack up indefinitely, and upgrades take effect immediately. The free trial is a one-time grant of 1,000 credits, enough to put every endpoint through its paces. The live numbers are on the pricing page.