Billing Platforms for AI Products Compared (2026)
For most AI products the choice comes down to one question: do you want a merchant of record (Paddle, Polar, Lemon Squeezy or Stripe Managed Payments) that handles global sales tax for roughly 5% per sale, or Stripe Billing at about 3.6% plus your own tax setup? Both can run free trials, subscriptions and usage-based billing for tokens or credits. Dedicated usage-billing engines (Metronome, which is now part of Stripe, plus Orb, Lago, Flexprice and Chargebee) make sense once you sell negotiated contracts, prepaid commits, credit wallets or complex metered pricing. All fees below come from each vendor's pricing page as of November 2026.
What an AI billing platform has to do
AI billing is harder than classic SaaS billing because your costs move with every request. A customer on a $20 plan can cost you $2 or $200 in inference depending on how they use it. A good AI billing platform handles four jobs that a plain subscription tool doesn't:
- Metering: ingest a usage event per request (tokens in and out, images, seconds of audio, agent runs) and aggregate it per customer, ideally in near real time.
- Rating: turn usage into money with per-unit, package, tiered or dimensional prices (for example, a different price per model or region).
- Credits and entitlements: prepaid credit wallets that burn down, free monthly allowances, top-ups, and hard limits that stop usage when the balance hits zero.
- Invoicing and collection: combine subscription fees, usage and one-time charges on one invoice, collect payment, handle tax, and give finance clean revenue data.
Plain AI subscription billing (one flat monthly plan) works until a few heavy users eat your margin. Most products then end up with hybrid pricing: a monthly platform fee that includes some usage, plus overages or credit packs. The best AI pricing for usage-based billing is usually the one your customers can predict, so plan for spend alerts and usage dashboards from day one.
Merchant of record vs payment processor
This decision matters more than the brand.
- Merchant of record (MoR): the platform is the legal seller. It calculates, collects and files sales tax and VAT worldwide, handles chargebacks, and answers customers' billing questions. You get a payout. Paddle, Polar, Lemon Squeezy and Stripe Managed Payments work this way.
- Payment processor plus billing (PSP): you are the seller. Stripe Billing, or a billing engine such as Chargebee, Orb, Metronome, Lago or Flexprice in front of a gateway, runs subscriptions and invoices, but registering for and filing taxes in each country is your job (Stripe Tax calculates and collects it, but you register in each jurisdiction and file, yourself or through a filing partner).
For a solo developer or small team selling an AI tool to consumers and small businesses worldwide, an MoR removes a lot of compliance work. For B2B products sold mostly to US companies, or once volume makes a 1.5-2 point fee difference meaningful, a PSP is cheaper.
Stripe Managed Payments says it takes over tax compliance in 80+ countries and 35 product categories, with Stripe (via Link) as the merchant of record (Stripe Managed Payments). That is the scale an MoR buys you: no registrations or filings on your side.
Billing platforms compared
The table covers the top generative AI billing platforms in 2026, from merchants of record to dedicated usage engines, with fees as published on each pricing page.
| Platform | Model | Fees (Nov 2026) | Usage-based / credits | Best for |
|---|---|---|---|---|
| Stripe Billing | PSP | 0.7% of billing volume + card processing (2.9% + 30¢ US cards) | Billing Meters API (100M events/month included); Stripe now points new usage integrations to Metronome | Teams that want full control and the lowest fees |
| Stripe Managed Payments | MoR | +3.5% per transaction on top of standard Stripe processing | Works with Stripe Billing subscriptions | Staying on Stripe while offloading tax |
| Paddle | MoR | 5% + 50¢ per transaction | Usage-based billing in early access (beta); custom line items and credit packs today | SaaS selling globally |
| Polar | MoR, open source (Apache-2.0) | Starter 5% + 50¢; Pro $20/mo + 3.8% + 40¢; Growth $100/mo + 3.6% + 35¢; Scale $400/mo + 3.4% + 30¢; +1.5% for non-US cards | Event ingestion, meters, metered prices, meter credits | Developer tools and AI apps that want usage billing built in |
| Lemon Squeezy | MoR (owned by Stripe) | 5% + 50¢, extra fees on some non-US payments | Usage records on subscriptions | Existing users; new projects should read its migration notice |
| Metronome | Usage-billing engine (a Stripe product) | Starter: 0.8% of billing volume + $0.04 per 1,000 events | Real-time metering, prepaid credits, commits, contracts, dimensional pricing | Usage-heavy AI APIs and enterprise contracts |
| Orb | Usage-billing engine | Business: $6 per credit (1 credit = 0.60% of billing volume, plus event ingestion); 100 free credits; Enterprise custom | Streaming ingestion, live alerts, hybrid and usage pricing | Larger companies with complex pricing |
| Lago | Usage-billing engine, open source (AGPL-3.0) | Self-host free; Cloud and Premium by quote | Billable metrics, events API, Kafka/Kinesis/S3 ingestion | Teams that want to self-host or own the code |
| Flexprice | Usage-billing engine, open source (AGPL-3.0) | $199/mo + 0.5% of billing (50M events/mo); Enterprise from $3,000/mo flat | Real-time metering, credit grants, top-ups, feature limits | AI startups that want credits and entitlements in one tool |
| Autumn | Billing and entitlements layer on Stripe | Free up to $8K monthly billing volume; Pro $375/mo; Stripe fees still apply | Credit balances, feature entitlements, usage enforcement | Shipping credits and paywalls fast on Stripe |
| Chargebee | Billing engine on 40+ gateways | Flow plan: 0.8% of invoicing volume, or $99/mo + 0.65% (100M usage events/mo included) | Usage ingestion, included usage, overages | Teams that want billing independent of one gateway |
What a $20/month subscription costs
For one $20 charge on a US card, before any tax-filing costs:
- Stripe Billing: $0.88 processing + $0.14 Billing = $1.02 (add $0.10 if you use Stripe Tax's 0.5% no-code option where you're registered)
- Stripe Managed Payments: $0.88 + $0.70 = $1.58 (plus Billing's $0.14 if the subscription runs on Stripe Billing)
- Paddle, Lemon Squeezy, Polar Starter: 5% + 50¢ = $1.50
- Polar Pro, Growth, Scale: $1.16, $1.07 and $0.98, plus their $20, $100 or $400 monthly fees
Non-US cards change this: Stripe adds 1.5% for international cards and 1% for currency conversion (Stripe pricing), and Polar adds 1.5% for non-US cards. Paddle's pricing page advertises no hidden extras on top of 5% + 50¢.
Merchant-of-record options (Paddle, Polar, Lemon Squeezy, Managed Payments) include global tax filing; Stripe Billing does not. Polar paid plans also charge a monthly fee. Source: Stripe, Paddle, Polar and Lemon Squeezy pricing pages
Polar's own breakeven math: Pro beats Starter at about $1,379 a month in sales, Growth at about $5,634, and Scale at about $19,048 (Polar pricing).
What a usage-billing engine costs at $50,000 a month
The engines charge on top of card processing, so compare their platform fee alone. Example: an AI API that bills $50,000 a month and sends 10 million usage events.
| Engine | How it charges | Platform fee at $50K billed + 10M events |
|---|---|---|
| Stripe Billing (Meters) | 0.7% of billing volume, events included up to 100M | $350 |
| Orb (Business) | $6 per credit, 1 credit = 0.60% of billing volume, plus event ingestion | $300 plus ingestion (not priced publicly) |
| Chargebee Flow | 0.8%, or $99 + 0.65% | $400 (the $99 option is $424; it wins above $66K) |
| Autumn Pro | $375/mo (50K billing volume included), plus Stripe's 0.7% | $725 |
| Flexprice | $199/mo + 0.5% of billing, 50M events included | $449 |
| Metronome Starter | 0.8% of billing + $0.04 per 1,000 events | $800 ($400 + $400) |
| Lago (self-hosted) | Free software, you pay for hosting and ops | Your infrastructure cost |
Excludes card processing and any Stripe Billing fee your setup adds. Orb's event ingestion charge is not published, so its bar is a floor. Source: Stripe, Orb, Chargebee, Autumn, Flexprice and Metronome pricing pages, November 2026
Price is rarely the deciding factor at this stage; the features you need (credits, commits, dimensional prices, self-hosting) usually decide for you. But the table shows why event-heavy products should check per-event pricing: at 100M events a month, Metronome's event charge alone would be $4,000.
The platforms in more detail
Stripe Billing
Stripe Billing handles subscriptions, invoices, proration, dunning and a hosted customer portal on top of Stripe Payments. Pay-as-you-go pricing is 0.7% of billing volume, including Billing transactions processed on and off Stripe and excluding one-off invoices. Annual plans (paid monthly, 1-year contract) come in tiers for up to $100,000, $250,000, $500,000 and $1,000,000 of monthly billing volume, with 0.67% on volume above the tier (Stripe Billing pricing). Basic usage-based billing via the Meters API is included, with up to 100 million events per month.
One change to know about: Stripe's usage-based billing guide now carries a "Not Recommended" banner for Billing Meters and points every new usage integration, including adding metering to existing flat-rate subscriptions, to Metronome, which Stripe owns (Stripe docs). Existing Meters integrations don't need to migrate, and Stripe says it continues to fully support them.
What Meters can't do, per Stripe's own comparison: prepaid credits with real-time drawdown, enterprise commits and minimums, ramp schedules, dimensional pricing, and real-time usage visibility (credits are reconciled only at invoice time, so customers can overspend during the cycle). What Metronome can't do yet: it has only limited Checkout support (custom API calls and webhooks), and no Connect, Adaptive Pricing, Workflows or Stripe Dashboard support (Stripe comparison).
A full Meters setup is three API calls. Create a meter that sums tokens per customer, a metered price, then send events:
curl https://api.stripe.com/v1/billing/meters \
-u "$STRIPE_SECRET_KEY:" \
-d "display_name=LLM tokens" \
-d event_name=llm_tokens \
-d "default_aggregation[formula]=sum" \
-d "customer_mapping[event_payload_key]=stripe_customer_id" \
-d "customer_mapping[type]=by_id" \
-d "value_settings[event_payload_key]=value"
# $0.04 per 100 tokens, billed monthly
curl https://api.stripe.com/v1/prices \
-u "$STRIPE_SECRET_KEY:" \
-d currency=usd -d unit_amount=4 -d billing_scheme=per_unit \
-d "transform_quantity[divide_by]=100" -d "transform_quantity[round]=up" \
-d "recurring[usage_type]=metered" -d "recurring[interval]=month" \
-d "recurring[meter]=$METER_ID" -d "product_data[name]=LLM tokens"
Stripe Managed Payments
Stripe's own MoR product for SaaS, digital content and virtual goods. It adds 3.5% per successful transaction to standard Stripe fees, and Stripe becomes the merchant of record, handling tax, fraud, disputes and customer support (Stripe Managed Payments). It's a good fit if you already built on Stripe Checkout and Billing and want to stop worrying about VAT. To see what Stripe keeps from a given price with Billing, Tax or Managed Payments switched on, use our Stripe fee calculator; Stripe's test card numbers cover declines, 3D Secure and disputes while you build.
Paddle
Paddle is a long-running MoR for software at a flat 5% + 50¢, including tax, fraud protection, chargebacks and buyer support, with no monthly or migration fees (Paddle pricing). For orders under $10, Paddle asks you to contact it for custom pricing, which matters for cheap AI credit packs.
The gap for AI products is narrowing. As of November 2026 Paddle's docs say native usage-based billing, including prepaid usage on subscription plans and per-customer balances, is in early access (beta) and you apply to join. Outside the beta you meter usage yourself (Paddle suggests a library such as OpenMeter) and add a custom line item to the next transaction, or sell credit packs as one-time products and track balances in your app (Paddle for AI companies).
Polar
Polar is an open source MoR aimed at developers (polarsource/polar, Apache-2.0, 10,460 stars). Its usage-based billing fits AI pricing directly: you send events such as ai_usage with a total_tokens field in their metadata, define a meter that sums them, attach a metered price to a subscription product, and grant monthly meter credits as a plan benefit (Polar docs). Metered prices work only on subscription products.
Pricing changed in 2026. Starter is free with 5% + 50¢ per transaction; Pro ($20/month), Growth ($100/month) and Scale ($400/month) lower the rate to 3.8% + 40¢, 3.6% + 35¢ and 3.4% + 30¢. Organizations created before May 27, 2026 keep an Early Member rate of 4% + 40¢ plus a 0.5% subscription fee until they change plans. Extras: +1.5% for non-US cards, $15 per dispute whatever the outcome, and Stripe's payout fees passed through at cost (Polar pricing). DevHunt listing: Polar.
Lemon Squeezy
Stripe acquired Lemon Squeezy in 2024. In its January 28, 2026 update, Lemon Squeezy said it is moving its merchant-of-record service onto Stripe Managed Payments, acknowledged slower support and fewer product updates during the transition, and said Managed Payments supported merchants in 35+ countries with access through a waitlist, with public sign-up "very soon" (Lemon Squeezy 2026 update). It still charges 5% + 50¢ and supports metered subscriptions through its usage records API, but for a new project, compare Stripe Managed Payments directly. DevHunt listing: Lemon Squeezy.
Metronome, Orb and Chargebee
These are billing engines, not processors. They take usage events, rate them against price books and contracts, and produce invoices that a gateway (usually Stripe) collects.
- Metronome is now part of Stripe; its pricing page says so directly, and Stripe's docs call it "Stripe's primary usage-based billing platform" and the recommended starting point for all new integrations. The Starter plan costs 0.8% of billing volume plus $0.04 per 1,000 ingested events; larger plans are custom (Metronome pricing). It supports pay-as-you-go, flat fee plus overage, credit-based pricing, prepaid credits with drawdown, commits and minimums, ramp schedules and dimensional pricing.
- Orb sells credits: on the Business plan each credit is $6 (less with volume discounts) and works out to 0.60% of billing volume, plus event ingestion and other metered dimensions. New accounts get 100 free credits. Business includes streaming event intake, live alerts, hybrid and usage-based billing, Orb invoicing, and Stripe and Adyen payment integrations; Enterprise adds customer hierarchies, NetSuite and Salesforce integrations and a 99.9% availability SLA (Orb pricing).
- Chargebee's Flow plan costs 0.8% of invoicing volume, or $99/month plus 0.65% (the fixed option wins above $66,000 a month), includes 100 million usage events a month, and connects to 40+ gateways (Chargebee pricing).
You need one of these when sales negotiates commits, minimums and ramp schedules, or when usage volume outgrows simple meters. Before then, it's extra cost and integration work.
Sending usage to Metronome is one HTTP call with up to 100 events per request; transaction_id is the idempotency key, so retries never double-bill (Metronome API):
curl -X POST https://api.metronome.com/v1/ingest \
-H "Authorization: Bearer $METRONOME_API_KEY" \
-H "Content-Type: application/json" \
-d '[{
"transaction_id": "req_8f3a2c",
"customer_id": "cust_42",
"event_type": "llm_request",
"timestamp": "2026-11-04T12:00:00Z",
"properties": { "model": "large", "input_tokens": 1200, "output_tokens": 350 }
}]'
Lago
Lago is the best-known open source usage billing engine (getlago/lago, AGPL-3.0, 10,756 stars, v1.56.0 released November 2, 2026). It turns events into billable metrics with six aggregations (COUNT, COUNT UNIQUE, LATEST, MAX, SUM, WEIGHTED SUM), combines them with plans for subscription, usage or hybrid pricing, and runs either as Lago Cloud or self-hosted (Lago docs). Lago does not publish Cloud prices; Premium features are sold by quote for both Cloud and self-hosted deployments (Lago pricing).
Events go to POST /api/v1/events, deduplicated by transaction_id:
curl -X POST "https://api.getlago.com/api/v1/events" \
-H "Authorization: Bearer $LAGO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"event": {
"transaction_id": "req_8f3a2c",
"external_subscription_id": "sub_42",
"code": "llm_tokens",
"properties": { "tokens": 1550, "model": "large" }
}
}'
The batch endpoint takes up to 100 events, and one invalid event rejects the whole batch with a 422. For higher volume, Lago also ingests from Kafka, Kinesis and S3, which are not subject to the REST rate limits (Lago ingestion guide). The AGPL license matters if you plan to modify Lago and offer it as a service; for internal use it rarely does.
Flexprice
Flexprice is a younger open source engine built for usage-based and credit-based pricing (flexprice/flexprice, AGPL-3.0, 6,874 stars, v2.1.47 released November 4, 2026). Its docs list real-time usage aggregation, programmatic prepaid and promotional credit grants with auto top-ups and expiry rules, feature limits per customer, seat-based, usage-based or hybrid plans, SDKs for JavaScript, Python and Go, and self-hosting (Flexprice docs).
Hosted pricing is $199/month plus 0.5% of billing with up to 50 million events a month; Enterprise starts at $3,000/month flat with no revenue share and adds VPC, on-prem or air-gapped deployment (Flexprice pricing). Events go to POST /v1/events with an x-api-key header and return 202 for asynchronous processing:
curl -X POST https://us.api.flexprice.io/v1/events \
-H "x-api-key: $FLEXPRICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"event_name": "llm_request",
"external_customer_id": "cust_42",
"event_id": "req_8f3a2c",
"properties": { "model": "large", "tokens": "1550" },
"timestamp": "2026-11-04T12:00:00Z"
}'
Note that Flexprice's API types properties values as strings, so send numbers as strings.
Autumn
Autumn sits between your app and Stripe and handles "subscription state, credit balances, feature entitlements, and usage enforcement" (Autumn docs). Instead of mapping Stripe webhooks to your own plan logic, you call Autumn's check before an action ("can this customer run another agent task?") and report usage after it. Your Stripe account stays in place, and Autumn says it is open source and can be self-hosted.
The Free plan covers up to $8K of monthly billing volume, 10,000 customers and 10M API requests; Pro is $375/month for $50K of monthly billing volume, usage event aggregation and multiple billing sources (Stripe, RevenueCat, Vercel); Scale is custom. Because Autumn is built on Stripe Billing, Stripe's 0.7% and 2.9% + 30¢ still apply (Autumn pricing). Pick it if the hard part for you is entitlements and credits, not invoicing.
Flexprice vs Lago for real-time metering (and Orb, Metronome)
These four come up together when an AI team wants a dedicated usage engine. The short version:
| Flexprice | Lago | Orb | Metronome | |
|---|---|---|---|---|
| Open source | Yes, AGPL-3.0 | Yes, AGPL-3.0 | No | No |
| Self-host / on-prem | Yes; Enterprise adds VPC, on-prem, air-gapped | Yes | No | No |
| Published price | $199/mo + 0.5% | Quote only (self-host free) | $6/credit = 0.6% of billing + ingestion | 0.8% + $0.04 per 1K events |
| Real-time | Real-time aggregation (docs) | Ingestion at scale, Kafka/Kinesis/S3 | Streaming intake, live alerts | Real-time usage tracking and credit burndown |
| Credits and limits (per vendor docs) | Credit grants, auto top-ups, expiry, feature limits | Plans plus billable metrics | Hybrid and usage pricing, live alerts | Prepaid credits, commits, real-time drawdown |
| Stripe relationship | Separate company | Separate company | Stripe and Adyen integrations | Owned by Stripe |
How to decide between Flexprice or Lago for metering events:
- Pick Lago if you want the more established open source project (more stars, longer release history) and plan to self-host at volume with Kafka or Kinesis ingestion.
- Pick Flexprice if credits, top-ups and feature limits are central to your AI product and you want a published hosted price, or need on-prem or air-gapped deployment with a flat Enterprise fee.
- Pick Metronome if you are on Stripe and selling enterprise contracts, since Stripe now builds its usage billing roadmap on it.
- Pick Orb if finance needs invoicing, receivables and NetSuite/Salesforce integrations from the billing system.
Flexprice alternatives, then, depend on what you want from it: Lago for open source, Autumn for entitlements on Stripe, Metronome or Orb for enterprise contracts, and Chargebee when you need many gateways.
Credit-based pricing software for AI platforms
Credits are the most common way AI products make usage predictable: customers buy or receive a balance, each request burns some of it, and the app stops or asks for a top-up at zero. What each option gives you:
| Option | Prepaid credits | Real-time drawdown | Auto top-up | Where the balance lives |
|---|---|---|---|---|
| Stripe Billing (Meters) | Credit-based pricing, no prepaid drawdown | No, reconciled at invoice time | Build it | Stripe |
| Metronome | Yes, plus commits | Yes | Yes (thresholds) | Metronome |
| Flexprice | Yes, credit grants | Real-time aggregation | Yes | Flexprice |
| Autumn | Yes, credit balances | Checked per action (check) |
Not stated | Autumn |
| Polar | Meter credits granted each billing cycle | Not stated | Not stated | Polar |
| Paddle | Credit packs as one-time products (native in beta) | Your app | Your app | Your app (or Paddle beta) |
Whatever generative AI usage billing software you pick, test the credit flow first. The practical rule: if a customer can burn through more inference than they paid for before the invoice is finalized, you need a real-time balance. That is exactly the case Stripe's own docs flag for Meters, where customers can exceed their balance during the cycle.
Managing credits and entitlements in AI agent applications
Agents make this sharper, because one user action can trigger dozens of model calls. Three patterns that work:
- Check, then act. Before each agent run, ask the billing layer whether the customer has balance or entitlement (Autumn's
checkcall, or a cached balance from Metronome or Flexprice). Deny or degrade gracefully when they don't. - Reserve, then settle. Hold an estimated cost when the run starts and settle the real token count when it ends, so long runs can't overdraw.
- Send one event per model call with the model name and token counts as properties. Dimensional pricing (Metronome) or event properties (Lago, Flexprice) then let you price models differently and still put everything on a single customer invoice.
Should you build billing in-house?
Building metering yourself is tempting: a usage table, a nightly job, and a Stripe invoice item. It breaks on the details: idempotent ingestion (retries that don't double-bill), late events, mid-cycle plan changes, credit expiry, proration, revenue reports finance trusts, and price changes without a deploy. Every engine above handles those, and a generative AI usage billing platform earns its fee on exactly these edge cases. A reasonable path:
- Start with Stripe Billing Meters or a merchant of record (Polar, Paddle) while pricing is simple.
- Add an engine (Metronome, Lago, Flexprice, Orb) when you need real-time credits, enterprise commits or dimensional prices.
- Keep your own usage log anyway. Store every event you send with its
transaction_id. When an invoice is disputed, you can prove the number.
Before you set a price per credit, run a few real prompts through DevHunt's token counter to see what each one costs you on each model.
What changed in 2026
- Jan 28, 2026Lemon Squeezy says it is moving its MoR service onto Stripe Managed Payments (update).
- May 27, 2026Polar's new plans apply to organizations created from this date; older ones keep the Early Member rate (pricing).
- By Nov 2026Metronome's site says it is now part of Stripe; Stripe's docs mark Billing Meters "Not Recommended" for new usage integrations (Stripe docs).
- By Nov 2026Paddle opens usage-based billing as an early-access beta (Paddle docs).
- Nov 2, 2026Lago v1.56.0 (releases).
- Nov 4, 2026Flexprice v2.1.47 (releases).
Set up a free trial that converts to paid (Stripe example)
AI products have a cost problem other SaaS doesn't: every trial user burns inference spend. Two common patterns:
- Time-limited trial with a usage cap. 7 or 14 days, but enforce a token or credit limit in your app so a single trial can't run up a large bill.
- Free credits, then pay as you go. Give a one-time credit grant (Polar meter credits, Metronome credits, or your own counter) and require a card to continue.
Here's a no-card 14-day trial with Stripe Checkout that cancels automatically if no payment method is added (Stripe docs):
import Stripe from 'stripe';
const stripe = new Stripe(process.env.STRIPE_SECRET_KEY);
const session = await stripe.checkout.sessions.create({
mode: 'subscription',
line_items: [{ price: process.env.PRO_PLAN_PRICE_ID, quantity: 1 }],
payment_method_collection: 'if_required', // no card at signup
subscription_data: {
trial_period_days: 14,
trial_settings: { end_behavior: { missing_payment_method: 'cancel' } }, // or 'pause'
},
success_url: 'https://example.com/welcome?session_id={CHECKOUT_SESSION_ID}',
});
// redirect the user to session.url
Then handle these webhooks:
customer.subscription.trial_will_end: sent 3 days before the trial ends by default. Email the user a link to add a card (Stripe's customer portal works for this). Stripe can also send its own reminder 7 days before, which card network rules require for some trials (trial events).customer.subscription.deleted(orcustomer.subscription.pausedif you chosepause): revoke access.invoice.paid: the user converted; grant the paid plan's limits.
Two limits to know: trials without a payment method don't support metered (usage-based) prices, so use a flat-price plan for the trial and add usage items after conversion. And Stripe's newer Trial Offer API (paid or discounted trials, API version 2026-09-30.endive) isn't supported in Checkout yet, so stick with trial_period_days when you use Checkout.
To report usage after conversion with Billing Meters:
await stripe.billing.meterEvents.create({
event_name: 'llm_tokens', // must match the meter's event name
payload: { stripe_customer_id: customerId, value: String(totalTokens) },
});
Meter events are processed asynchronously, so a preview of the upcoming invoice may not include the latest events yet. Price per token only after you know your own model costs; DevHunt's AI model pricing table compares API prices across providers.
Which billing platform should you choose?
- Solo developer, selling worldwide, flat plans: Paddle or Polar. If you need token metering from day one, Polar.
- Already on Stripe and want tax handled: Stripe Managed Payments.
- US-focused B2B, or volume where fees matter: Stripe Billing plus Stripe Tax, and register for sales tax where you cross thresholds.
- Usage-heavy API with enterprise deals: Metronome (or Orb) in front of Stripe.
- Open source or on-prem required: Lago or Flexprice.
- Credits and feature gating on Stripe without building it: Autumn.
- Multiple gateways or regions: Chargebee.
Migrating later is painful because saved cards are tied to the processor. Pick something you can live with for a couple of years. If you're incorporating through Stripe Atlas, check the Stripe credits listed on DevHunt's perks page, and browse more billing tools in payments tools.
FAQ
What is the best billing platform for AI products?
For small teams, a merchant of record with usage billing built in, such as Polar, or Paddle if your plans are flat. For larger usage-based products, Stripe with Metronome, which Stripe now recommends for new usage-based integrations. The right answer depends mostly on whether you want to handle sales tax yourself.
How does usage-based billing work for AI tokens?
Your app sends an event per request (customer ID plus tokens used) to the billing platform. A meter sums the events per billing period, a metered price converts the total into money (for example $0.04 per 100 tokens), and the platform adds it to the next invoice. Prepaid credits work the same way but draw down a balance first.
Is Stripe a merchant of record?
Standard Stripe Billing is not: you're the seller and responsible for tax. Stripe Managed Payments is Stripe's merchant-of-record option, at 3.5% on top of standard processing fees as of November 2026.
Is Lemon Squeezy still available?
Yes, but it's in transition. Stripe owns it, and Lemon Squeezy said in January 2026 that its merchant-of-record service is moving to Stripe Managed Payments. Existing merchants can keep selling; check its latest update before starting a new project on it.
What does usage-based billing cost on Stripe?
Basic metering through the Meters API is included in Stripe Billing's 0.7% fee, up to 100 million events per month. Metronome, Stripe's advanced usage billing product, publishes a Starter plan at 0.8% of billing volume plus $0.04 per 1,000 events.
What is the best billing platform for AI startups with usage-based pricing?
Start with Polar (if you want a merchant of record) or Stripe Billing Meters (if you don't), and move to Metronome, Lago, Flexprice or Orb when you need real-time credits or enterprise contracts. Those are the best platforms for AI usage-based billing at their respective stages; none is best at every stage.
Is Metronome part of Stripe now?
Yes. Metronome's pricing page says "Metronome is now part of Stripe," and Stripe's docs describe it as Stripe's primary usage-based billing platform. It still has its own pricing (0.8% plus $0.04 per 1,000 events on Starter) and its own API.
Flexprice or Lago for metering events?
Both are open source (AGPL-3.0) and self-hostable. Lago is older and more widely adopted, with Kafka, Kinesis and S3 ingestion for high volume. Flexprice focuses on credits, top-ups and feature limits and publishes a hosted price ($199/month plus 0.5%). For real-time metering both aggregate continuously; test with your own event volume before committing.
Which monetization platform should AI founders use for billing based on tokens and API calls?
Send one event per API call with token counts as properties, then pick the engine by stage: Polar or Stripe Meters for a first launch, Metronome or Flexprice when you need real-time credit balances, Orb or Chargebee when finance needs enterprise invoicing.
Which usage-based billing platform has the strongest developer API and documentation?
Judge it yourself by sending a test event in each sandbox. On published docs alone: Stripe's are the most complete, Metronome documents idempotent batch ingestion of up to 100 events per call, and Lago and Flexprice are open source, so you can read the code behind every endpoint.
Should generative AI platforms deploy specialized usage billing infrastructure or adapt traditional subscription billing engines?
Adapt a subscription engine while pricing is a flat plan with a usage cap; Stripe Billing Meters or a merchant of record covers that. Move to specialized usage billing infrastructure (Metronome, Orb, Lago, Flexprice) once you need real-time credit balances, per-model pricing or enterprise commits, because subscription engines reconcile usage only at invoice time.
Can I add usage-based pricing without altering existing core application code?
Mostly. The billing engine owns plans, prices and invoices, so price changes need no deploy. Your app still has to send one usage event per billable action (a single HTTP call or SDK line in your API middleware) and, for credits, check the balance before serving a request.
How do I aggregate variable multi-model compute costs onto a single customer invoice?
Send every model call as an event with the model name as a property. Use dimensional pricing (Metronome) or property-based metrics (Lago, Flexprice) to apply a different rate per model, and the engine sums them into one invoice line per model on the same customer invoice.
Are there free usage-based billing platforms?
Yes. Lago and Flexprice are free to self-host, Autumn is free up to $8K of monthly billing volume, Polar's Starter plan has no monthly fee (you pay per transaction), and Orb gives 100 free credits to test.