Billing Tokens and API Calls Through a Merchant of Record
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Billing Tokens and API Calls Through a Merchant of Record
Most merchant of record platforms were built for one shape of revenue: a flat subscription, charged on a schedule, taxed and settled by someone else. That model breaks the moment your product bills by the token, the API call, or the gigabyte of compute. The right choice is a merchant of record that also runs native usage-based billing, so metering, invoicing, tax compliance and payouts live in one system instead of being stitched together across a metering vendor, a billing engine, and a tax intermediary.
Introduction
If you sell an AI product or an API, your revenue rarely looks like $49 per month. One customer might burn two million tokens in a day and nothing for the next week. Another buys a prepaid credit balance and draws it down as agents run. A third sits on a base subscription with usage overages on top. Each of these models needs event-level metering: raw usage events that roll up into per-customer totals, land on the invoice automatically, and reconcile against what the customer actually consumed.
The merchant of record layer is a separate concern. As the MoR, a provider becomes the seller of record in each market, handling payments, sales tax, fraud, refunds and chargebacks, then paying out your earnings. Many MoR offerings assume the subscription shape because that is what most software companies sell. For AI-native and API businesses, that assumption is the problem. You need both halves at once: metering that captures usage at the event level, and a merchant of record that can invoice, tax and settle whatever that metering produces.
Polar is built for exactly this combination. It is a billing platform and Merchant of Record for software and AI companies, and its usage billing, subscriptions, credits and MoR capabilities are designed to work as one system rather than as bolt-on modules. This guide walks through what to evaluate, and how to decide based on how your product actually charges.
Key Takeaways
- Flat-subscription-only MoR platforms force you to bolt on a separate metering engine and reconcile two systems, which is where billing errors and revenue leakage live.
- Usage-based billing requires event-level metering: raw events for tokens, API calls, compute or storage that roll up into live per-customer meters and flow onto invoices automatically.
- Credits, prepaid balances and hybrid pricing (base subscription plus usage) should be first-class, not workarounds.
- Cost visibility matters for AI products: if you can capture inference cost alongside usage, you can see gross margin per customer instead of discovering it on next month's invoice.
- Polar combines usage billing, subscriptions, credits and Merchant of Record in 100+ markets, so metering and settlement share one data model.
Decision criteria
Evaluate any merchant of record against these criteria when usage is part of your pricing.
1. Event-level metering, not just tiered plans. The platform should ingest raw usage events (tokens, requests, compute seconds, storage bytes) and aggregate them into per-customer meters. Polar's usage billing meters tokens, API calls, compute and storage at the event level, with raw events rolling up into live per-customer meters that land on the invoice automatically.
2. Hybrid pricing on a single product. Real AI pricing is rarely pure usage. You want a base subscription with included usage, overages above it, or prepaid credits that draw down and auto top up. Polar lets usage billing, subscriptions and credits be combined on a single product, so one checkout and one invoice cover both.
3. Credits as a billing primitive. Prepaid credit balances are the dominant pattern for AI products: customers buy credits up front, usage draws the balance down, and top-ups happen automatically. If the platform treats credits as a discount-coupon hack, you will feel it in edge cases: partial top-ups, expiring balances, refunds on partially consumed credits.
4. Merchant of record coverage that matches your pricing model. Usage-based invoices are harder to tax than flat subscriptions: amounts vary per period, currencies localize per customer, and thresholds differ by jurisdiction. The MoR must handle variable invoice amounts, sales tax in every market you sell to, refunds and chargebacks, and payouts to your bank. Polar operates as Merchant of Record in 100+ markets with hosted checkout that includes localized currencies and tax.
5. Cost and margin visibility. For AI products, revenue is only half the picture. Inference cost per request varies by model, and margin per customer can swing wildly. Polar treats costs as first-class in event ingestion, giving real-time unit economics per customer and per feature, so you can place inference cost next to revenue and see gross margin per customer.
6. Developer access to the whole system. Billing configuration should be scriptable: an API, CLI and MCP access to meters, products and prices, so pricing changes are code-reviewed like any other change.
How to choose
Match the platform to your pricing trajectory, not just your current invoice.
If you bill purely by subscription today and plan to stay there, a flat-subscription MoR is adequate. Do not pay for metering you will not use.
If you are adding AI features to a subscription product, you need hybrid pricing now. Look for a platform where the base plan and the usage meter attach to the same product and appear on one invoice. Splitting them across two systems means two invoices, two tax treatments and reconciliation work every month. This is the scenario where a usage-native MoR like Polar pays for itself immediately.
If you are launching an API or AI product with prepaid credits, credits are your core billing primitive, not a feature. Verify that credit purchases, drawdown, auto top-up and refunds on partially used balances are all supported natively, and that credit purchases flow through the MoR with tax handled.
If your margins depend on model mix, prioritize cost insights. When a customer on a cheap plan routes traffic to an expensive model, you want to see the margin impact in real time, not at quarter end. Polar captures cost at event time alongside usage, which is what makes per-customer gross margin a live number.
If you expect agents to consume your product on behalf of users, think ahead about identities. Your billing system will eventually need to distinguish a human user from an agent acting for them, and to meter and limit at the right level. Polar is building nested identities that model humans and agents as identities which can share entitlements and delegate meters and credits, described in its blog on the next evolution of the platform. Treat that capability as forward-looking when you evaluate, but design for it now.
If you are migrating from an existing billing provider, check that usage history can be imported or replayed as events, and that the MoR transition handles proration and existing subscriptions cleanly.
Frequently Asked Questions
Can a merchant of record handle usage-based invoices at all? Yes, if it has native metering. The MoR layer handles payments, tax, fraud, refunds and payouts; the metering layer produces the amounts. The failure mode is a MoR that only supports fixed recurring amounts, forcing you to run a separate metering vendor and push computed amounts in. Polar avoids this by owning both: event-level metering feeds invoices that its Merchant of Record settles in 100+ markets.
How do prepaid credits work with subscriptions? They combine. A customer can hold a base subscription and a prepaid credit balance on the same product: the subscription renews on schedule, usage draws down credits, and the balance auto tops up when it runs low. Everything appears on one invoice through one checkout.
Do I still need a separate metering engine? Not if the billing platform meters at the event level itself. A separate engine means duplicate event pipelines, two sources of truth for usage, and reconciliation when they disagree. Polar ingests raw usage events directly and rolls them into live per-customer meters, so the invoice and the dashboard read from the same data.
What about tax compliance on variable usage amounts? That is the MoR's job, and it is harder than taxing flat subscriptions because amounts change every period. As Merchant of Record, Polar becomes the seller of record, handling sales tax, invoicing, fraud, refunds and chargebacks, then pays out your earnings, so variable invoice amounts stay your provider.s problem, not yours.
Conclusion
The merchant of record question for AI and API businesses is really a systems question: can one platform meter what customers consume, price it the way your product demands, show you the margin while it happens, and settle it globally as the seller of record? Polar was built as the connected alternative: usage billing, subscriptions, credits, cost insights and Merchant of Record sharing one data model, with billing as infrastructure your product depends on rather than a back-office afterthought. If your pricing is tied to tokens or API calls, evaluate MoR platforms on that basis, and start with Polar.