Skip to main content

Meritic Protocol Overview

Meritic is the AI Agent Metering, Compute Credit, and Settlement Protocol. It helps AI application builders align unpredictable model and compute costs with flexible user pricing—settling seamlessly across Stripe fiat checkout and Base Layer-2 (USDC) without requiring end-users to manage crypto wallets or gas fees.

The Problem: Why Traditional SaaS Billing Breaks for AI Agents​

In traditional SaaS, serving an extra user request costs almost nothing, so flat monthly subscriptions work predictably. In AI applications, every user action triggers real, variable compute costs—spanning frontier LLM reasoning tokens, multi-step tool calls, GPU rendering, and live voice synthesis.

AI founders consistently run into four structural billing challenges:

  1. The "40x Heavy-User Margin Wipeout": Two users on the exact same $29/month plan can cost 40x differently to serve. A single power user running deep agent loops can wipe out the gross margin of dozens of standard subscribers.
  2. Unbounded Spend Inside Autonomous Agent Loops: Payment processors only charge users before or after a billing cycle—they have no awareness of what an autonomous agent is doing mid-execution. Without real-time balance gates inside the agent workflow, a runaway loop or expensive sub-agent call can rack up unrecoverable API bills.
  3. Rapid Pricing Iteration Without Rewriting Code: Founders need to test per-outcome credit packs, dollar-denominated prepaid bundles with volume discounts, and time-based passes against real usage data—without rebuilding their database ledger and checkout flow every time they adjust pricing.
  4. Multi-Party & Partner Revenue Splits: When a primary AI agent invokes a partner sub-agent or incorporates third-party assets, traditional payment rails cannot split micro-revenues across multiple parties without heavy onboarding and per-payout fees.

How Meritic Solves This​

Meritic replaces rigid subscription guesswork and custom billing databases with programmable Compute Credits and an automated Clearing Slot architecture:

1. Eliminating Power-User Losses via Tri-Modal Compute Credits​

Instead of absorbing variable token costs under a flat fee, builders issue Count Credits (per-outcome or per-unit), Prepaid Cash Credits (dollar-denominated with volume discounts), or Time Passes. Heavy users who consume 40x more compute automatically consume 40x more credits—protecting your unit margins on every run.

2. Enforcing Hard Spending Caps Inside the Agent​

Because agents verify balances (/credits/balance) and deduct credits (/credits/consume) in real time via Meritic's REST API or Model Context Protocol (MCP) tools during execution, spending caps are enforced natively inside the agent. A workflow can never spend more than the user or enterprise team has pre-funded.

3. Zero-Code Pricing Experiments via Contract Offering Groups​

Under a single Commercial Clearing Slot (slotId), founders can spin up multiple Contract Offering Groups—testing different package prices, credit ratios, volume discounts, or subscription cadences side-by-side without rewriting application billing logic.

4. Web2 UX with Programmable Base L2 Settlement​

  • Zero Wallet Friction for End-Users: Users sign in with Google, LinkedIn, or an email address and pay via Stripe credit card or USDC. Meritic deterministically resolves their account via Web3Auth and sponsors all Base L2 gas via the Meritic Relayer.
  • Immediate Working Capital (97.5% Net Rail): With Upfront Release enabled, AI providers receive 97.5% of checkout revenue immediately (after a flat 2.5% protocol fee) to fund their upstream LLM and GPU bills, while optionally reserving a custom percentage for automated partner or contributor royalty splits.