Usage-based billing software in 2026: compare pricing and features

Usage-based billing software splits into four architecture groups: subscription-first (Stripe Billing), invoice-based (Metronome, Orb, Lago), real-time (Credyt, Stigg), and hybrid (Flexprice). Subscription-first and invoice-based platforms record usage during a cycle and bill at cycle end; real-time platforms authorize and debit the customer's balance at the moment of usage. This article ranks all seven on architecture, pricing, and fit.

At a glance: ranking

The ranking below reflects fit for products with variable per-request infrastructure costs (AI, API, GPU). Different rankings apply for different angles; Stripe Billing sits at #7 here because subscription-first architecture is the weakest fit for variable per-request costs, not because it is the weakest billing platform overall. See what usage-based billing is for the underlying concept.

Usage-based billing software: feature comparison

The table below compares all seven platforms on 10 concrete dimensions. The rows that matter most for AI product teams are architecture (real-time vs post-usage), usage authorization (pre-usage vs post-usage), and wallet primitive (first-class vs add-on).

Pricing at the canonical scenario

At a 100-customer, $2,000/month scenario, Credyt costs $90/month, Stripe Billing costs ~$72 plus payment processing, and Stigg costs $448/month at the Growth floor. Flexprice ranges $500 to $1,000/month depending on tier; Metronome, Orb, and Lago do not publish pricing and require a sales conversation. The scenario uses 100 customers, $20 ARPU, 1,000 events per customer per month, 5 seats, $2,000 monthly revenue, and 100,000 monthly events; the same parameters apply to every platform so the costs are directly comparable.

Pricing notes:

• Credyt: $90 = 90 active wallets × $1/MAW, with the first 10 wallets free.
• Stigg: per-unit costs total ~$193, but the Growth plan has a $448/month annual contract minimum, so that is the effective cost.
• Metronome: pricing is not published; enterprise quotes require a sales conversation. Payment collection runs through the merchant's PSP at PSP rates. Source: Metronome website, accessed May 2026.
• Orb: no longer publishes pricing. Estimate based on previously public rates captured by third-party analysis; may not reflect current rates.
• Lago: cloud Business and Enterprise tiers both require sales contact as of April 2026. The self-hosted AGPLv3 core is free to run but does not include the customer portal, credit notes, automatic dunning, tax integrations, CRM connectors, email invoices, or Lago AI agents. Source: Lago pricing, accessed May 2026.
• Flexprice: Starter is $500/month; Premium is $1,000/month. The vendor does not publicly confirm whether real-time prepaid credit balance, recurring wallet top-ups, and entitlement management are available at the Starter tier. If Premium is required for those, the effective comparison cost is $1,000. Annual billing reduces these by 20%. Source: Flexprice pricing, accessed May 2026.
• Stripe Billing: Starter at 0.5% covers the canonical scenario; a Scale-plan customer would pay $16 (0.8% × $2,000) before processing. Payment processing (2.9% + $0.30 on US cards) is separate and card-mix dependent. Source: Stripe Billing pricing, accessed May 2026.

Why architecture matters

Usage-based billing splits on a single architectural axis: when does billing settle. Subscription-first and invoice-based platforms record usage events during a cycle and produce an invoice at cycle end; real-time platforms authorize the action against the customer's balance and debit immediately as the event arrives. Hybrid platforms straddle both modes.

The choice is not stylistic. It dictates what the platform can do.

Real-time architectures can block a runaway agent before its inference cost lands; invoice-based architectures discover the overrun at cycle end. Real-time architectures expose a current customer balance for self-service top-up; invoice-based architectures expose an invoice that arrives later. Real-time architectures hold tokens or GPU hours as native assets; invoice-based architectures convert custom units to USD before invoicing.

For products with variable per-request infrastructure costs, the difference is concrete. Frontier model inference is priced in dollars per million input and output tokens, which translates to cents to dollars per request depending on context length and model choice (Anthropic API pricing, accessed May 2026). If a customer fires 50 requests in a minute on Claude, the team has paid the inference cost in real time. Stripe completed its acquisition of Metronome on January 14, 2026 (Stripe newsroom) specifically to add real-time metering and enterprise contract capabilities to Stripe Billing; this is a direct market signal that the architecture gap exists.

The seven usage-based billing software options

The seven cards below are ordered by fit for products with variable per-request infrastructure costs, from highest architectural alignment to lowest. Each card uses the same five-block format: positioning, best-for, pricing, three strengths, two trade-offs.

1. Credyt

Real-time billing infrastructure for AI products with variable per-request, per-token, or per-inference costs. Customers prepay into multi-asset wallets; usage is authorized before it occurs and debited immediately as a single atomic operation.

Best for: Real-time billing for AI products with variable per-request costs.

Pricing: $0 to start (10 free wallets); $1 per Monthly Active Wallet in production; first 1M events per month free; no revenue percentage; no markup on PSP fees. Source: Credyt pricing, accessed May 2026.

Strengths:

• Real-time per-usage authorization: customer balance is checked before the AI action runs, so spend cannot exceed budget.
• Multi-asset wallets: USD, tokens, GPU hours, and custom units in a single customer wallet, each with its own grants and lifecycle.
• Drop-in branded customer portal with customer-controlled auto top-up; no frontend engineering required.

Trade-offs:

• Designed for prepaid wallet and real-time billing; not the right fit for high-volume enterprise contract billing with quarterly reconciliation.
• Smaller partner ecosystem and brand recognition than Stripe Billing.

Learn more: real-time billing without replacing the existing stack.

2. Stigg

Monetization control layer that sits between a product and a downstream billing system (Stripe, Zuora, Chargebee). Real-time entitlement checks and credit consumption run inside Stigg; payment collection and invoicing run downstream at cycle end.

Best for: Real-time entitlements and credit orchestration over an existing billing system.

Pricing: Sandbox $0; Growth from $448/month annually with $0.32 per self-service subscription; Scale custom. Source: Stigg pricing, accessed May 2026.

Strengths:

• getMeteredEntitlement() API gates feature access in real time with sub-100ms P95 latency (Stigg docs, accessed May 2026).
• No-code pricing console for non-engineers; draft and publish plan versioning ships pricing changes without code deploys.
• Billing-system agnostic: orchestrates Stripe, Zuora, Chargebee, AWS Marketplace, Apple App Store.

Trade-offs:

• Stigg is real-time for entitlements but cycle-end for billing; payment collection runs through the connected system.
• Growth plan has a $448/month contract minimum even when per-unit costs come in lower.

Learn more: Credyt vs Stigg head-to-head.

3. Metronome

Enterprise-grade usage-based billing with a SQL-based rating engine, multi-year contracts, commitments, and 34-day backdating. Streams billions of usage events monthly for OpenAI, Anthropic, Databricks, and NVIDIA. Acquired by Stripe; the deal closed January 14, 2026, and integration is in flight.

Best for: High-volume enterprise metering with custom contracts.

Pricing: Not published; enterprise quotes require a sales conversation. Source: Metronome website, accessed May 2026.

Strengths:

• SQL-based billable metrics: custom aggregation logic without bespoke engineering.
• Enterprise contract management: multi-year contracts, amendments, true-ups, multi-product deals.
• Multi-motion GTM: unified self-serve, sales-led enterprise, and cloud marketplace billing.

Trade-offs:

• Post-usage invoicing model; usage accumulates and is billed at cycle end. No real-time wallet debit primitive.
• 34-day backdating window constrains retroactive corrections and pricing changes.
• Customer portal requires integration via signed URLs; no drop-in self-service portal.

Learn more: Metronome documentation.

4. Orb

Revenue platform for high-volume usage-based billing with custom SQL metrics, dimensional pricing, pricing simulation against historical usage, and hosted rollups for millions of events per second. Customers include Vercel, Replit, Supabase, Neo4j, and LaunchDarkly.

Best for: Enterprise usage billing with custom SQL metrics and dimensional pricing.

Pricing: Not publicly listed; enterprise quotes require a sales conversation. Third-party analysis estimates ~$720/month at the canonical scenario; figure may not reflect current rates. Source: Orb website, accessed May 2026.

Strengths:

• Custom SQL metrics for complex aggregation logic without custom engineering.
• Pricing simulation: validate changes against historical usage before deploying.
• Hosted rollups for millions of events per second; deep finance integrations (NetSuite, Salesforce, data warehouses).

Trade-offs:

• Post-usage invoicing model; usage accumulates and is billed at cycle end. Threshold invoicing triggers mid-cycle but is still post-consumption.
• Customer portal is pre-authenticated and invoice-oriented; no self-service balance or top-up.
• Pricing requires sales contact; no public price list as of May 2026.

Learn more: Credyt vs Orb head-to-head.

5. Lago

Open-source AGPLv3 core for usage-based, subscription, and hybrid billing. 9,500+ GitHub stars (Lago GitHub, accessed May 2026); named customers include Mistral AI, Groq, Synthesia, and PayPal. Available as self-hosted or managed cloud.

Best for: Open-source self-hosted invoice-based billing for engineering-led teams.

Pricing: Self-hosted AGPLv3 core is free; cloud Business and Enterprise tiers require sales contact. Source: Lago pricing, accessed May 2026.

Strengths:

• AGPLv3 core: forkable, self-hostable, auditable; SOC 2 Type II certified.
• Supports subscriptions, usage-based, prepaid credits, coupons, and entitlements in a single install.
• Active community and development; integrates with Stripe, Adyen, GoCardless, Salesforce, NetSuite, Xero.

Trade-offs:

• Customer portal, credit notes and refunds, automatic dunning, tax integrations, and email invoices are paid-tier features; the free core does not include them.
• Wallet balance updates at invoice finalization, not on event ingestion (Lago wallet docs, accessed May 2026); no pre-usage authorization primitive.
• Self-hosting a production deployment requires Postgres, Redis, workers, observability, and ongoing engineering for upgrades.

Learn more: Credyt vs Lago head-to-head.

6. Flexprice

Open-source AGPLv3 monetization infrastructure with a no-code pricing dashboard and ClickHouse-backed metering. Targets early-stage SaaS and AI teams adopting metered or hybrid pricing for the first time.

Best for: Hybrid billing for early-stage SaaS adopting usage-based pricing.

Pricing: Free trial (3 months); Starter $500/month (10M events); Premium $1,000/month (25M events). Source: Flexprice pricing, accessed May 2026.

Strengths:

• Open-source AGPLv3 core, self-hostable on Docker Compose or AWS ECS/EKS.
• No-code pricing dashboard; non-engineers change pricing without deployments.
• Wallet primitives with auto top-up, low-balance alerts, and promotional credit grants.

Trade-offs:

• Usage-based wallet debit is invoice-driven, not event-driven: charges accumulate and debit on INVOICE_PAYMENT, not as events arrive (Flexprice wallet transactions docs, accessed May 2026).
• Customer portal is read-only; auto top-up is merchant-controlled, not customer-self-service.
• PSP support is limited to Stripe, Razorpay, Paddle, and Chargebee; no bring-your-own PSP.

Learn more: Credyt vs Flexprice head-to-head.

7. Stripe Billing

Stripe's subscription and usage-based billing product, tightly integrated with Stripe's payments, tax, and revenue recognition stack. Strong defaults for SaaS with global tax, ASC 606 revenue recognition, and Smart Retries. Now part of the same family as Metronome.

Best for: Subscription-first SaaS billing with metered add-ons.

Pricing: Starter at 0.5% of recurring payments; Scale at 0.8%; Custom for enterprise. Stripe Payments processing (2.9% + $0.30 on US cards) is separate. Source: Stripe Billing pricing, accessed May 2026.

Strengths:

• Full payments stack on one account: processing, fraud, tax, compliance, and revenue recognition.
• Global payment methods: 50+ methods and 135+ currencies with local acquiring.
• Mature SDKs, libraries, documentation, and third-party integrations; large developer ecosystem.

Trade-offs:

• Subscription-first architecture: subscriptions and usage-based plans accumulate charges through billing cycles and generate an invoice at cycle end. No real-time wallet debit primitive.
• Customer Portal is invoice-centric; does not expose a real-time prepaid balance or self-service wallet top-up.
• Tightly coupled to Stripe Payments; replacing or mixing PSPs means rebuilding billing.

Learn more: Credyt vs Stripe Billing head-to-head.

How to choose

If you bill per AI inference, per token, or per request and need to authorize spend before the model runs: Credyt is the best fit. Real-time per-usage authorization is the architecture that blocks runaway spend before the cost lands; invoice-based platforms can alert but cannot prevent the charge.

If you keep Stripe, Zuora, or Chargebee as your settlement layer but need real-time entitlement and credit logic in the product: Stigg is the best fit. Stigg orchestrates entitlements in real time and lets billing settle downstream.

If you have an enterprise sales motion with multi-year contracts, custom SQL pricing, and millions of usage events per second: Metronome or Orb is the best fit. Both are purpose-built for high-volume invoice-based billing with deep contract machinery.

If your team needs to self-host billing infrastructure for compliance, data residency, or vendor lock-in reasons: Lago is the best fit. The AGPLv3 core is self-hostable; budget for the engineering cost of running production infrastructure.

If you are an early-stage SaaS team adopting usage-based pricing for the first time and want a no-code pricing dashboard: Flexprice is the best fit. The platform targets teams moving from flat subscriptions to metered billing without dedicated billing engineering.

If your billing is primarily subscription with occasional metered add-ons and you already run payments on Stripe: Stripe Billing is the best fit. AI teams looking for alternatives to Stripe Billing for variable per-request costs evaluate the platforms above.

Bottom line: choosing usage-based billing software

Usage-based billing software splits into four architecture groups: subscription-first (Stripe Billing), invoice-based (Metronome, Orb, Lago), real-time (Credyt, Stigg), and hybrid (Flexprice). The architecture choice is a fit choice, not a hierarchy; match it to whether usage cost lands in real time or aggregates over a cycle, then choose by feature within that branch. For AI products where every inference has direct infrastructure cost, Credyt's approach to real-time per-usage authorization is purpose-built for the case.

Frequently asked questions

What is the difference between subscription-first, invoice-based, and real-time billing?

Subscription-first platforms (Stripe Billing) are designed for recurring subscriptions with metered add-ons; usage is a secondary primitive. Invoice-based platforms (Metronome, Orb, Lago) capture events in a metering layer during a cycle and reconcile them into an invoice at cycle end. Real-time platforms (Credyt, Stigg) authorize and bill against the customer's balance at the moment of usage as a single atomic operation. The difference matters most for products with variable per-request infrastructure cost.

Why was Stripe's acquisition of Metronome significant for the usage-based billing market?

Stripe acquired Metronome to fill a gap in Stripe Billing: high-throughput metering, SQL-based billable metrics, and enterprise contract management. The deal closed January 14, 2026 (Stripe newsroom). Combined product roadmap is in flight as of May 2026; specific GA dates for shared features are not public.

Is Stigg a Stripe Billing replacement or a layer on top of it?

Stigg is a layer on top, not a replacement. It orchestrates entitlements, credit consumption, and pricing changes in real time, while payment collection and invoicing run through the connected billing system (Stripe, Zuora, or Chargebee) at cycle end. Teams that want to keep Stripe Billing as the settlement layer can add Stigg for real-time entitlement and credit logic in the product.

Can Lago authorize usage in real time?

No. Lago meters events in real time but the authoritative wallet balance updates at invoice finalization, not on event ingestion. The premium ongoing-balance estimate refreshes every five minutes (Lago wallet docs, accessed May 2026). There is no documented primitive to query the wallet and block a specific action before it runs.

What does Orb cost at small scale?

Orb does not publish pricing. Third-party analysis of previously public rates suggests ~$720/month at the canonical 100-customer scenario; the figure may not reflect current rates. Pricing requires a sales conversation. For the same scenario, Credyt is $90/month with public pricing.

When does open-source billing infrastructure matter for the architecture choice?

Open source is a deployment choice, not an architecture choice. Lago is invoice-based and Flexprice is hybrid, despite both being AGPLv3. Open source explains who can self-host the codebase; it does not explain how billing settles.

Teams choosing open-source billing should evaluate the architecture separately and account for the engineering cost of running production infrastructure (Postgres, Redis, workers, webhooks, PSP wiring) alongside the license benefit. For implementation patterns once the architecture is chosen, see consumption-based pricing implementation.

Which platform fits AI products with variable per-request costs?

Real-time end-to-end fits best. Credyt is purpose-built for AI products with per-inference, per-token, or per-action costs that hit before revenue is collected. Stigg is a viable option for teams keeping a downstream settlement layer (Stripe, Zuora, Chargebee) and adding real-time entitlement and credit logic on top. Invoice-based platforms can issue alerts but cannot block usage before the cost lands; for an AI product where a runaway agent can burn through a budget in minutes, that distinction is the architecture choice.