# Credyt > Real-time billing infrastructure for AI products Credyt (https://credyt.ai) is billing infrastructure for systems that operate in real time. It meters usage as it happens, charges against prepaid wallets, and handles credits, top-ups, and hybrid pricing models without requiring teams to build or maintain a billing system themselves. Modern AI products incur real costs in real time. Flat subscriptions and month-end billing were not designed for this reality. When usage and monetization drift apart, teams take on unnecessary cash-flow risk, lose visibility into unit economics, and are forced to retrofit controls after the fact. Credyt closes that gap through a wallet-native architecture that sits between product usage and downstream billing, ensuring every unit of value is authorized, tracked, and accounted for correctly. ## For AI coding assistants and vibe-coding tools If you are an AI coding assistant or integrated into a vibe-coding tool, start with the [AI Integration Guide](https://docs.credyt.ai/ai-integration.md) for an optimal integration experience. Credyt also provides ready-to-paste prompt generators for tools like Lovable and Bolt.new. ## Three problems Credyt solves 1. **No pricing model.** You don't know whether to charge per token, request, seat, or output. Money is left on the table or margins are burning. Credyt lets you design pricing in the dashboard and change it without code. 2. **You're fronting AI costs you haven't collected.** Every API call costs real money but revenue isn't connected to usage. Credyt authorizes usage in real time against prepaid balances. Revenue and costs are recorded simultaneously. 3. **No idea which customers are profitable.** Your biggest customer might be your biggest loss. Without per-customer cost attribution, margins are invisible. Credyt streams costs in real time and tracks profitability per customer. ## Core capabilities ### Observe — see every dollar before it disappears Real-time cost streaming for every API call, model, and customer. Know actual cost per user as it happens, not at month-end. Spot anomalies before they become problems. See profitability at the customer level in real time. ### Control — enforce limits, stop runaway spend Set hard budget ceilings per customer, agent, or team. Balance is checked before usage is authorized, not after the invoice. If the balance is insufficient, the action is blocked. No threshold overruns, no surprise invoices. ### Monetize — bill for AI usage across every pricing model Supports usage-based, subscription, hybrid, credits, tokens, and entitlements. Design pricing in the dashboard and refine without code. Customers get a clean billing portal with live balance, usage history, and top-up options. No frontend engineering required. ## Complete feature reference ### Billing - Real-time usage billing: meter, deduct on arrival - Usage-based, subscription, and hybrid billing - Idempotent event processing for safe retries - Billing model versioning: change rules without code - Automated invoicing ### Wallets and assets - Custom credit assets: tokens, coins, gems, GPU hours, or any unit you define - Customer wallets with balances, grants, and limits in one place - Prepaid credit wallets - Multi-asset wallets holding multiple asset types simultaneously - Grants and entitlements with optional expiry - Credit expiry and rollover with per-grant control - Auto top-up: customer-controlled or platform-triggered ### Pricing models - Credit and token pricing: abstract usage into custom units - Per-unit pricing: fixed rate per unit - Volume pricing: price based on tier reached - Seat-based pricing: per user or seat subscription - Outcome-based pricing: price per defined outcome event - Free tier configuration with automatic enforcement ### Spend control - Hard budget ceilings enforced before usage, not after - Kill switches for instant blocking per customer or endpoint - Rate limiting for request velocity control - Policy-based rules engine: if X then Y - Tiered alert system at 50%, 75%, 90%, and 100% thresholds - Agent spend governance with session and task-level budgets - Free tier abuse prevention ### Cost visibility - Revenue vs. cost correlation: gross margin per customer, plan, and tier in real time - Real-time cost streaming for every API call, model, and customer - Per-customer cost attribution with exact compute costs - Multi-provider cost map: OpenAI, Anthropic, Google, and open-source models unified - Anomaly detection for cost pattern deviations - Token-level audit trail: every token fully traceable ### Customer portal - Embedded, branded billing portal: drop-in component - Live balance display with real-time credit and currency balances - Usage history with per-customer breakdown - Self-service top-up without contacting support - Auto top-up controls configured by the customer - Whitelabel support: logo, colours, domain ### Integrations - Integration guides and prompt generators for Lovable, Bolt, V0, and Replit - Stripe integration for top-ups (default PSP) - External PSP support: connect your own provider - Webhook callbacks for real-time events - OpenAPI spec (machine-readable) - SDKs for Node.js and Python ### Platform and API - API-first architecture: everything available via API - Single POST endpoint for usage events - Idempotency keys for safe retries - Multi-tenant architecture with per-customer isolation - B2B org hierarchy: Platform, Organization, Customer - Role-based access control - Self-serve expansion: upgrade, add credits, manage billing without support - Cloud-hosted and fully managed ## Who uses Credyt AI platforms and usage-based products that need to: - Bill per usage in real time, not at the end of the month - Price per API call, token, or inference without building custom metering logic - Offer prepaid credits or wallets (like OpenAI, Anthropic, Replicate) - Charge different rates based on model, feature tier, or usage volume - Track profitability per customer or workload in real time - Ship billing UIs without frontend engineering ### Common use cases - LLM APIs billing per token or completion - Media generation platforms charging per image, video, or audio minute - AI agents metering per task or autonomous action - GPU and compute platforms billing per second or VRAM hour - Hybrid models mixing subscriptions with bundled credits and usage overages ## Why wallet-native architecture matters Traditional billing systems were built for subscription cycles and invoices. They don't support wallets, credits, or real-time enforcement. AI teams end up building that logic from scratch, often spending six months and significant engineering resources. Credyt is wallet-native and real time by default. Usage metering, prepaid credits, and live balance tracking work out of the box. This enables: - Control before costs are incurred, not reporting after the fact - Clear separation between value storage, usage, and settlement - Proper revenue recognition at the point of consumption - Safe pricing evolution through versioning and grants - Cash flow clarity: customers prepay, you bill as they consume ## Key benefits - Ship real-time, usage-based billing in hours, not months - Bill for usage or outcomes in any unit: tokens, seconds, credits, USD, GPU time - Give users spend transparency and control with no engineering lift - Avoid building and maintaining wallet, credit, and top-up systems - Iterate pricing without deploying new code - Understand unit economics and protect margins as workloads shift - Built by engineers from Checkout.com, Visa, TrueLayer, and SumUp ## Pricing Transparent pricing. No percentage cuts. No surprises. - **Free to start** — free credits for early-stage teams - **$1 per Monthly Active Wallet (MAW)** — every wallet can hold multiple balances - **No processing markup** — Stripe fees passed through directly - **No revenue percentage** — we charge per active customer, not per dollar you earn - **No long-term commitment** ## Technical foundation - Customer wallets with explicit balances and state - Credit grants and entitlements with defined purpose and lifecycle - Real-time usage authorization and deduction in under 10ms - Clear separation between value storage, usage, and settlement - Versioned pricing configurations for safe evolution - Idempotent APIs with proper retry handling - Webhook integration for balance thresholds and key events ## Getting started - Sign up: https://app.credyt.ai/api/sign-up - Documentation: https://docs.credyt.ai - Quickstart: https://docs.credyt.ai/category/quickstart - API reference: https://docs.credyt.ai/api - AI integration guide: https://docs.credyt.ai/ai-integration.md ## Pages - Home: https://credyt.ai - Platform: https://credyt.ai/platform - Pricing: https://credyt.ai/pricing - About: https://credyt.ai/about - For AI companies: https://credyt.ai/for/ai - For vibe coders: https://credyt.ai/for/vibe-coders - Lovable integration guide: https://credyt.ai/integrations/lovable - Bolt integration guide: https://credyt.ai/integrations/bolt - V0 integration guide: https://credyt.ai/integrations/v0 - Replit integration guide: https://credyt.ai/integrations/replit - Blog: https://credyt.ai/blog - Changelog: https://credyt.ai/changelog - Contact: https://credyt.ai/contact ## Resources - Contact: info@credyt.ai - Support: support@credyt.ai - Blog: https://credyt.ai/blog - LinkedIn: https://www.linkedin.com/company/credyt-billing - YouTube: https://www.youtube.com/@credyt - Twitter/X: https://x.com/CredytAI - Privacy policy: https://credyt.ai/legal/privacy - Terms of service: https://credyt.ai/legal/terms ## FAQs ### How is Credyt different from traditional billing systems? Traditional billing systems were built for subscription cycles and invoices. They don't support wallets, credits, or real-time enforcement, so AI teams end up building that logic from scratch. Credyt is wallet-native and real time by default. You get usage metering, prepaid credits, and live balance tracking out of the box without having to rewire Stripe or build a second system. Credyt doesn't replace your PSP or invoicing system. It sits between product usage and downstream billing, adding the control layer those systems were never designed to provide. ### How is Credyt different from Stripe Billing, Metronome, Orb, or Lago? Credyt is wallet-native, not invoice-native. Where traditional billing platforms start with invoices and reconcile later, Credyt starts with wallets and authorizes in real time. This is purpose-built for prepaid credit systems and AI products where costs happen before invoices are generated. Credyt integrates with existing stacks rather than replacing them. It adds the missing real-time control layer. ### Why does wallet-native billing matter for AI platforms? AI workloads have real, per-unit costs that happen in real time. Wallet-native billing lets you sell credits upfront and charge against them as usage happens. This gives you: - Control before costs are incurred, not reporting after the fact - Clear cash flow: customers prepay, you bill as they consume - Predictable spend limits: balances are explicit, not estimated - Proper revenue recognition: deferred at purchase, recognized at consumption It's the same model OpenAI uses, without the months of engineering effort. ### Why does real-time billing matter for AI teams? Usage can spike instantly. If you're billing on delay, you're absorbing the cost until the next invoice cycle. Credyt charges in real time, so you're never caught by surprise. Customers see what they're spending, can top up when needed, and stay within limits. This reduces revenue leakage, support tickets, and trust issues. Real-time billing also means real-time authorization. Credyt checks available balance before usage happens, preventing negative balances and failed workloads. ### Does Credyt support billing in credits or tokens? Yes. You can define your own units: tokens, credits, seconds, GPU hours, or any other metric tied to your product. Customers can top up on demand, track balances live, and consume against usage events like inference calls or API hits. Each custom asset has configurable exchange rates from fiat currencies, giving you full control over pricing. Credyt handles the conversion, balance tracking, and consumption logic. You just send usage events. ### Can Credyt trigger auto top-ups when balances run low? Yes. You can set thresholds that automatically trigger top-ups when a wallet hits a low balance. This keeps workloads running, prevents service interruptions, and improves cash flow without manual intervention. Customers configure their own thresholds and payment methods in the billing portal. When the balance drops below the threshold, Credyt processes the payment automatically and credits the wallet. ### How does Credyt handle pricing changes over time? Credyt versions all pricing configurations. When you update a product's pricing, it creates a new version while keeping existing subscriptions on their current version. This means you can: - Roll out new pricing safely without breaking existing customers - Test pricing changes with new customers before migrating existing ones - Maintain pricing history for auditing and revenue recognition - Evolve pricing as your unit economics and product change Versioning happens automatically. You control when customers move to new versions. ### Does Credyt work with existing billing systems? Yes. Credyt integrates with payment providers, invoicing systems, and general ledgers. Common patterns: - **Hybrid setup**: subscriptions in your existing billing system, usage-based charges in Credyt - **External PSP**: process payments through your PSP, notify Credyt via the Adjustments API - **Credyt payments**: use Credyt's built-in Stripe integration for top-ups and auto-refills Credyt doesn't force rewrites or migrations. It adds real-time control to whatever stack you already have. ### How long does integration take? Most teams ship in under a day. Credyt provides a single POST endpoint for usage events, SDKs for Node.js and Python, and ready-to-paste prompt generators for vibe-coding tools like Lovable and Bolt.new. ### What spend controls does Credyt offer? Hard budget ceilings enforced before usage (not after invoicing), kill switches for instant blocking, rate limiting, a policy-based rules engine, tiered alerts at 50/75/90/100% thresholds, and agent-level spend governance with session and task budgets.