TL;DR
- The best tool depends on which problem hurts most: metering depth (Orb, Metronome), owning the infrastructure (Lago, OpenMeter), staying on payment rails you already use (Stripe Billing), or getting metering, payments, and tax in one platform (Commet).
- Most usage-based billing engines are not Merchants of Record — with them you still assemble a payment processor, a tax solution, and your own compliance.
- Commet is the only tool on this list that combines native usage-based billing (metered, credits, balance) with Merchant of Record payments and tax handling.
- If you are deciding how to implement the billing itself, the usage-based billing for AI guide covers consumption models and metering patterns step by step.
| Tool | Usage metering | MoR / tax handling | Credits & balance | Open source | Pricing model |
|---|---|---|---|---|---|
| Commet | Native: metered, credits, balance | Yes — Merchant of Record | First-class credit packs + balance wallets | SDKs only | 4.5% + $0.40 per successful transaction, no monthly fees |
| Stripe Billing | Metered prices, you report usage | No — you are the seller | No native credit system | No | 0.7% of billing volume on top of processing fees |
| Orb | Advanced: SQL aggregation, backfill | No | Prepaid credits | No | Custom enterprise contracts |
| Metronome | Enterprise-scale, real-time | No | Credits and commitments | No | Custom enterprise contracts |
| Lago | Flexible aggregation types | No | Prepaid credits via wallets | Yes | Free self-hosted; cloud free to $250K cumulative, then 0.75% |
| OpenMeter | Real-time, metering-first | No | Entitlements and credit grants | Yes | Free self-hosted; managed cloud plans |
1. Commet — usage-based billing and Merchant of Record in one
Commet is a billing platform built for SaaS and AI products that charge by consumption. It is the only tool on this list that handles both halves of the problem: metering usage and legally selling the product.
Where it shines. Usage is a first-class primitive, not an add-on. Plans support three consumption models — metered with overage, prepaid credits, and balance wallets — plus seat-based pricing. For AI products specifically, Commet maintains a catalog of 180+ AI models with current token prices, so you set a margin per feature instead of hardcoding rates per provider. And because Commet is a Merchant of Record, tax calculation, collection, and remittance are handled for you: customers pay by card in their local currency in 20+ markets, and payouts arrive in local currency in 112 countries.
Integration is two calls — create a client, track an event:
import { Commet } from "@commet/node";
const commet = new Commet({ apiKey: process.env.COMMET_API_KEY! });
await commet.usage.track({
customerId: "user_123",
feature: "ai_chat",
model: "gpt-4o",
inputTokens: 1500,
outputTokens: 300,
});Where it falls short. Metering is less configurable than the specialists below: Commet covers the common patterns (count, sum, overage, token pricing) but does not offer SQL-based aggregation pipelines or multi-dimensional event transforms. Payments are card-only — no PayPal, wire transfers, or local payment methods today — and Commet does not issue locally certified e-invoices in countries that mandate them.
Pricing. A flat 4.5% + $0.40 per successful transaction, no monthly fees or minimums. That includes payment processing, tax handling, and all MoR services — see pricing for details.
2. Stripe Billing — usage billing on the payment rails you already have
Stripe Billing is the billing layer on top of Stripe's payment infrastructure. If you already process payments with Stripe, it is the path of least resistance.
Where it shines. Ubiquity and ecosystem. The documentation is excellent, every framework has an integration, and metered prices let you bill per unit of usage on top of subscriptions. For simple per-unit metering attached to an existing Stripe account, it works without adding a vendor.
Where it falls short. Usage-based billing is possible but manual: you track usage yourself, report it back correctly, reconcile invoices, and handle mid-cycle changes. There is no native credit or balance system, so prepaid AI pricing means building your own ledger. Webhooks are mandatory, and the sync layer between Stripe and your database is where bugs live. You also remain the merchant of record — tax registration and remittance stay your responsibility. The Stripe alternative comparison breaks this down in depth.
Pricing. Stripe Billing charges 0.7% of billing volume on top of standard payment processing fees, with additional fees for advanced features.
3. Orb — the deepest metering engine
Orb is a usage-based billing specialist built for complex metering and aggregation.
Where it shines. Event pipelines. Orb handles ingestion, deduplication, SQL-based aggregation, backfill and retroactive amendment, and flexible pricing models (per-unit, tiered, bulk, package, matrix). If your billing logic requires joining multiple event streams or applying custom SQL transforms, Orb is purpose-built for it. Strong choice for enterprises with genuinely complex usage models.
Where it falls short. Orb solves metering, not selling. It does not collect payments, handle taxes, or act as Merchant of Record — you pair it with a payment processor and a separate tax solution, and you maintain that integration. There is no real-time entitlement layer either; access control stays in your application. The full breakdown is in the Orb alternative comparison.
Pricing. Custom enterprise contracts negotiated on volume. There are no public prices, which adds friction for startups that want to start small.
4. Metronome — usage billing at enterprise event scale
Metronome is a usage-based billing platform aimed at large infrastructure and AI companies with very high event volumes.
Where it shines. Scale and real-time aggregation. Metronome is designed to ingest enormous event streams and keep balances, commitments, and credits current in real time — the pattern large AI API providers use for prepaid commitments and enterprise contracts. Credits, commitments, and custom contract terms are well supported.
Where it falls short. Like Orb, Metronome is a metering and rating engine, not a Merchant of Record: payments, tax, and compliance remain your problem and your vendor list. The product is built around enterprise sales — contract negotiation and implementation support rather than self-serve onboarding — which makes it a heavy lift for an early-stage product.
Pricing. Custom enterprise contracts; pricing is not public.
5. Lago — open-source billing you run yourself
Lago is open-source billing infrastructure you can self-host, with a managed cloud option.
Where it shines. Control and transparency. You see every line of code, own your data completely, and pay nothing for the software if you self-host. Aggregation is flexible, with multiple aggregation types and filters, and prepaid credits work through wallets. Lago has a strong developer community and a transparent roadmap.
Where it falls short. You own the infrastructure: deployment, uptime, backups, upgrades, and scaling are on you. Lago does not process payments — you connect your own processor and manage failures, retries, and reconciliation — and it does not handle taxes, so registration, remittance, and filings stay with you.
Pricing. Free to self-host. Lago Cloud is free for the first $250K in cumulative billing, then 0.75% of revenue, with a Performance plan at $599/month.
6. OpenMeter — open-source metering built for AI workloads
OpenMeter is an open-source, metering-first project focused on real-time usage collection for AI and developer-tool products, with a managed cloud offering.
Where it shines. Real-time metering as a primitive. OpenMeter is built to collect and aggregate high-frequency usage events — tokens, API calls, compute time — with low latency, and exposes entitlements and credit grants on top. Being open source, you can self-host it and inspect exactly how aggregation works.
Where it falls short. OpenMeter starts from metering, and the billing layer around it (invoicing, subscriptions) is younger than the rest of this list. It does not process payments, handle tax, or act as Merchant of Record, so it is one component in a stack rather than a complete billing system.
Pricing. Free to self-host under an open-source license; the managed cloud offers paid plans priced on usage.
How to choose
AI API startup (usage is the product). Your revenue is tokens and requests, and you have no billing stack yet. Assembling metering + payments + tax as three vendors is the slowest possible start. Commet covers the full path — track events, sell credits or balance, invoice, collect, remit tax — with transaction pricing that scales from zero. If you would rather own the infrastructure and have the engineering time, self-hosted Lago or OpenMeter gets you metering without license costs, but payments and tax remain yours to solve.
SaaS adding AI features. You already have subscriptions; you need a metered AI feature on top without a pricing migration. If you are on Stripe Billing and the feature is simple per-unit metering, staying put is reasonable. If you want credits or balance for the AI feature, real-time entitlements, or you are tired of webhook reconciliation, a platform with native consumption models saves you the ledger you would otherwise build. The usage-based billing for AI guide shows the add-a-metered-feature pattern concretely.
Enterprise with complex usage models. If your metering requires SQL transforms over multiple event streams, retroactive backfill, or contract-specific rating at massive event volumes, Orb and Metronome are the specialists — that depth is what their enterprise contracts buy. You will still pair them with payment processing and a tax solution, which at enterprise scale you likely have already.
The honest summary: every tool on this list is good at the job it was designed for. The question is how much of the billing problem you want one tool to own — just the metering, or the metering plus the selling.