How you charge for your product is as important as what your product does. AI companies are shipping tools that charge for tokens, API calls, compute minutes, agent runtime, and time spent inside a feature. Flat-rate subscriptions cannot map cleanly to any of that.
This piece covers the most relevant usage-based billing platforms in 2026, what each one is genuinely good at, and what to weigh before choosing.
What separates good billing infrastructure from bad
Before the platform list, here are the criteria that actually matter in a usage-based context.
Metering granularity. Can the platform ingest events at the atomic level (tokens, API calls, seconds of compute) without data loss or deduplication errors? Systems that aggregate at the wrong level create invoice disputes and revenue leakage.
Pricing model flexibility. Flat rate, per-seat, volume tiers, pay-as-you-go, prepaid credits, hybrid. The platform should handle all of these without requiring a separate integration for each model.
Merchant of Record coverage. If you sell globally, you need a platform that handles VAT, GST, and sales tax compliance. MOR coverage removes significant legal and operational overhead from your team.
Developer experience. How long does it take to go from sign-up to a live billing flow? The answer should be measured in hours, not weeks.
Transaction cost transparency. All-in fees vary dramatically across platforms. The headline rate is never the full story.
The platforms
Metronome
Metronome, now a Stripe product, is built for infrastructure companies with complex enterprise billing requirements: committed spend contracts, credit drawdowns, contract-level overrides, and billing at hundreds of millions of events per month. ElevenLabs, Anthropic, and Confluent run on it.
The tradeoff is complexity and cost. Metronome is custom-priced and sales-driven. For teams that need enterprise contract management, it is the reference implementation. For teams that do not, it is overbuilt.
Best for: Enterprise AI infrastructure companies with multi-year contracts and high event volumes.
Orb
Orb is developer-first, cloud-only, and designed for teams that want real-time customer-facing usage dashboards without building the metering layer from scratch. It uses SQL-based pricing metrics, which gives product teams genuine flexibility to define billing logic without writing custom billing code.
Pricing starts around $1,000 per month, making it a meaningful commitment for early-stage teams. The developer experience is strong and the documentation is thorough.
Best for: High-growth API companies that want real-time usage visibility and are ready to invest in dedicated billing infrastructure.
Lago
Lago is open-source (AGPLv3), self-hosted or cloud, and built for teams that want full billing ownership without vendor lock-in. It handles event ingestion, flexible pricing models (graduated, percentage, package, volume), and invoice generation.
Self-hosting is free. The cloud-managed version is priced separately. For teams with the engineering capacity to run their own infrastructure, Lago offers the strongest combination of flexibility and cost control. The tradeoff is operational overhead.
Best for: Teams that want open-source billing control and are comfortable owning the infrastructure.
Stripe Billing
Stripe Billing is the default choice for teams already on Stripe. It handles subscriptions, invoices, and basic usage-based pricing reasonably well. The integration path is familiar and the documentation is extensive.
The problem is cost at scale. Stripe charges 2.9% + $0.30 on transactions, plus 0.7% for subscription billing, 1.5% for international cards, and 1% for currency conversion. All-in, you are looking at approximately 9.6% + $0.30 per transaction for international billing. For a product charging $50 per month, that is $4.80 per transaction going to fees.
Stripe also does not act as Merchant of Record by default. Tax compliance and global payment liabilities stay on your plate unless you add Stripe Tax separately, and that is another system to maintain.
Best for: Teams in the earliest stages that are already on Stripe and need simple subscription billing before their pricing model gets more complex.
Flexprice
Flexprice is a hybrid billing platform built explicitly for AI pricing models. It supports real-time metering, token-based billing, prepaid credits, and multi-dimensional pricing in a single system. It targets AI-native teams that need more flexibility than Stripe Billing offers but are not ready for the operational overhead of Metronome or Orb.
Best for: AI companies that need hybrid usage-based pricing (credits, tokens, API calls) and want a purpose-built billing layer without enterprise-level commitments.
tiun
tiun sits in a different category from the other platforms on this list. It is not a dedicated billing engine you bolt on top of your stack. It is a unified commercial backend that handles authentication, payments, billing, customer data, and analytics through a single SDK.
Where Metronome and Orb require you to already have an auth provider (Clerk, Auth0), a payment processor (Stripe), and a customer database, and then wire them together with webhook logic, tiun replaces that entire layer. There are no webhook endpoints to maintain, no reconciliation jobs, no user tables to manage.
For billing specifically, tiun supports two product types: subscriptions and time-based billing. Time-based billing charges for time spent within your product rather than a flat membership rate, which is a genuine primitive for AI products that charge by session time, agent runtime, or content access windows. Both product types run through the same SDK call.
tiun also acts as Merchant of Record by default, so global tax compliance is handled without an additional integration. Integration happens through an SDK and optionally through an MCP server, which means AI coding agents can handle the setup work directly. The dashboard manages product creation, and the SDK takes over checkout, access control, and session management from there.
Where tiun differs concretely is in what it eliminates from the stack. Most early-stage AI products end up running three separate services before their first dollar of revenue: an auth provider, a payment processor, and some form of customer database. Each service sends webhooks. Each webhook listener needs to be written, deployed, kept in sync, and debugged when it drifts. When a payment succeeds, the auth database needs to know. When a subscription lapses, access needs to be revoked. That sync logic is not difficult to write once, but it is tedious to maintain, and it fails in the specific quiet ways that create support tickets and refund requests.
tiun handles that entire surface through one SDK. Checkout, access control, session gating, and analytics all run through the same call chain. There are no separate webhook endpoints to maintain because the state transitions happen inside the platform rather than across platform boundaries. For teams building on top of AI coding agents, tiun also exposes an MCP server, which means the integration work itself can be handled by an agent rather than requiring manual SDK wiring.
On cost, tiun's all-in transaction fee sits at approximately 3.4% + $0.30, which is the lowest of any MOR-inclusive platform in this category.
Best for: AI and SaaS companies at the early-to-growth stage that want to replace multiple point solutions with a single backend that acts as Merchant of Record and keeps transaction costs low.
How to choose
The right platform depends on your stage and your billing complexity.
Early stage, moving fast. If you want your commercial infrastructure handled without building a backend or managing webhooks between multiple systems, tiun covers auth, billing, and analytics in one place with MOR included and the lowest all-in fees in this category.
Dedicated usage metering at scale. If metering hundreds of millions of events per billing cycle is your core problem, Orb handles this well once you have graduated from early stage. It is a focused tool that does its job seriously.
Enterprise contracts. If you are selling to large enterprises with committed spend, annual true-ups, and contract complexity, Metronome is purpose-built for that. It is not the appropriate choice before you reach that scale.
Open-source control. If you want full billing infrastructure ownership and are comfortable with operational complexity, Lago is the strongest option in that category.
Already on Stripe, simple pricing. If your pricing is straightforward subscriptions and you are not yet selling internationally at volume, Stripe Billing is a serviceable starting point. Be clear-eyed about the fee structure before you scale.
The question underneath the platform question
Usage-based billing is table stakes for AI products in 2026. The real question is how much infrastructure you want to own.
Every system you add to your stack is a system you maintain, reconcile, and debug. At $10K MRR, the overhead is manageable. At $500K MRR, webhook failures at 2 a.m. and fee structures that compound across currencies are meaningful problems. The tools that compress that surface internally give engineering teams back the time they would otherwise spend on sync logic and payment reconciliation.
Billing infrastructure is not a commodity decision. The fees you pay today are the margins you keep or lose as you scale.
Frequently asked questions
What is usage-based billing and how does it differ from subscription billing?
Subscription billing charges a fixed amount on a recurring schedule regardless of how much the customer uses the product. Usage-based billing ties the charge to actual consumption: API calls made, tokens processed, compute minutes used, or time spent inside a feature. In practice, most AI and SaaS products end up with a hybrid of both: a base subscription that covers access, plus a usage layer that scales with the customer's actual activity. The billing platform you choose needs to handle both without requiring a separate integration for each model.
What does Merchant of Record mean and why does it matter for SaaS companies?
A Merchant of Record (MOR) is the legal entity that takes responsibility for the transaction: collecting payment, remitting taxes, handling chargebacks, and managing local payment regulations. When a platform is your MOR, you are not responsible for calculating and remitting VAT in Germany, GST in Australia, or sales tax across US states. That compliance burden transfers to the platform. For SaaS companies selling globally, operating without an MOR means either building significant tax compliance infrastructure internally or limiting which markets you sell into. Most early-stage teams cannot afford either option.
When should an AI startup move off Stripe Billing to a dedicated usage-based billing platform?
The signal is usually one of three things: your pricing model now involves dimensions that Stripe Billing cannot express cleanly (token counts, credit drawdowns, time-based access), you are losing revenue to fee structures that compound against you at scale, or your engineering team is spending meaningful time on webhook sync logic between Stripe and the rest of your stack. Any one of those is a reasonable trigger to evaluate alternatives. The cost question in particular tends to arrive earlier than teams expect: at $50K MRR with significant international volume, the difference between a 3.4% all-in rate and a 9.6% rate is more than $3,000 per month.
How do usage-based billing platforms handle failed payments and dunning?
Dunning refers to the automated process of retrying failed payments and notifying customers when their card declines or expires. Most dedicated billing platforms handle this natively: they retry on a configurable schedule, send reminder emails, and pause or revoke access after a defined number of failures. The quality of dunning logic matters more than most teams realize. A platform with aggressive, well-timed retries recovers a meaningful percentage of failed payments that would otherwise churn. The main variation across platforms is how configurable the retry cadence is and whether access revocation is tied directly to billing state or requires custom webhook logic to implement.