Introduction

What is Paid?

Paid is a billing platform purpose-built for businesses that use AI. Whether you’re building AI agents, integrating LLMs into your product, or offering AI-powered services, Paid helps you track costs, measure usage, and bill customers based on actual value delivered.

Core Capabilities

  • Usage-Based Billing: Charge customers based on usage, outcomes, seats, or any combination—not just flat subscriptions.
  • Invoicing: Automatically generate professional invoices with detailed usage breakdowns.
  • Pricing Plans: Create flexible pricing plans with support for tiered, volume, and graduated pricing.
  • AI Telemetry: Get observability over your AI workflows with real-time cost tracking across providers.
  • Cost Tracking: Track costs automatically from OpenAI, Anthropic, Mistral, ElevenLabs, and 50+ providers, or log costs manually.
  • Visualization: Use Blocks to create embeddable dashboards that show customers their usage and value in real-time.
  • Credits: Support prepaid credits and entitlement-based access.
  • Multi-Currency: Bill customers in their preferred currency.

Key Primitives

Paid is built around a few core concepts:

  • Products

    Products represent the AI-powered offerings you sell—chatbots, AI assistants, analytics tools, or any service that performs work. Each product can have its own pricing attributes and billing configuration.

  • Signals

    Signals are discrete usage events generated when your products perform work. Each time your AI sends a message, processes a request, or completes a task, a Signal captures that activity. Signals are the foundation for usage-based billing.

  • Customers

    Customers are the businesses or individuals you bill. Paid manages customer accounts, payment methods, entitlements, and billing history.

  • Orders

    Orders represent billing agreements with customers. An order contains line items that tie products to specific pricing, and tracks what has been billed.

  • Plans

    Plans are packaged offerings that bundle products and pricing together. Customers can subscribe to plans, and Paid handles upgrades, downgrades, and proration automatically.

  • Invoices

    Invoices are generated automatically based on usage and orders. They provide itemized breakdowns so customers see exactly what they’re paying for.

  • Blocks

    Blocks are embeddable visualization components that let customers see their usage, costs, and value delivered in real-time dashboards.

How It Works

  1. Define Your Products

    Create products in Paid that represent your AI offerings. Configure pricing attributes for each billable activity or outcome.

  2. Stream Signals

    Integrate Paid’s SDK or API to stream Signals whenever your products perform work. Signals capture what happened, when, and for which customer.

  3. Track Costs

    Paid automatically tracks your costs from AI providers, giving you real-time visibility into margins and profitability.

  4. Bill Customers

    Set up pricing models that match your business—usage-based, seat-based, outcome-based, or hybrid. Paid aggregates Signals and generates invoices automatically.

  5. Show Value

    Use Blocks to give customers transparency into what they’re getting. Demonstrate ROI with usage breakdowns and cost comparisons.

Flexible Billing Models

Paid supports the billing models that work for AI businesses:

  • Setup Fee (One-Time): Charge when a customer first signs up or onboards.
  • Platform Fee (Recurring): Regular fees for ongoing access.
  • Seat-Based (Recurring): Charge per user or seat.
  • Activity-Based (Usage): Charge per action—per message, per API call, per task completed.
  • Outcome-Based (Usage): Charge for results—per meeting booked, per lead qualified, per ticket resolved.
  • Hybrid Models: Combine fixed fees with usage-based charges for predictable base revenue with upside.

Integration

Paid provides SDKs for Python, Node.js, Go, and Ruby, plus OpenTelemetry integration for capturing telemetry from your existing observability stack. Most teams integrate in under 48 hours.

You can also ingest telemetry from external sources like Datadog and use it to define Signals and extract usage data.