AI cost per user tools compared

Tracking what each user costs in AI is the first half. Knowing whether that user is profitable needs the second half: what they actually paid. Here is how the tools people ask us about handle both. Facts were checked on 2026-11-02 from each vendor's public site; prices change, so check theirs too. We make FiscaLLM, so read this with that in mind, and tell us if anything is wrong.

ToolPriceOpen sourceRevenue for margin
FiscaLLMFree. Hosted with fair use (100,000 events per app per month), or self-hosted with no limitYes, MIT (server, dashboard and SDKs)Real revenue from RevenueCat (one-click connect and import), Stripe (webhook and 12-month import, beta, USD), CSV upload, or a revenue API
WeckrHobby free (50,000 requests a month), Pro $49 a month (1M requests a month)SDKs are MIT; the service is hostedMargin uses plan prices you set in the SDK; its site does not mention importing payments from Stripe or RevenueCat
MarginDashHosted plans; see its pricing pageNoStripe, or revenue events sent through its SDK
MarginMeterSee its siteNot statedStripe balance transactions, net of refunds, disputes and tax, through a read-only key
LangfuseFree self-hosted; hosted plans with a free tierYesTracks cost per user ID, but does not import subscription revenue to compute margin
HeliconeFree tier; paid hosted plansYesTracks cost per user through request headers, but does not import subscription revenue to compute margin

How to choose

See FiscaLLM's demo dashboard or start free.