# Pay.ml

> The Payment Layer for AI Agents.

Pay.ml is a financial technology platform that provides the payment layer for autonomous AI systems. Developers use it to issue secure, policy-bounded virtual cards for machine-to-machine (M2M) transactions — giving AI agents the ability to pay for APIs, cloud services, and digital goods without exposing primary credit lines or risking runaway costs.

## Core Problem

Traditional payment rails are built for human buyers. Autonomous AI agents face three distinct challenges:

1. **The Loop Risk**: An agent stuck in an infinite execution loop can drain a credit line in minutes. Traditional cards have no per-call or session-level spending controls.
2. **The Authorization Gap**: Credit cards restrict *how much* is spent, not *what* is bought. Agents need semantic, context-aware authorization policies expressed in plain English.
3. **The Integration Hurdle**: Manually sharing API keys or card details with untrusted agent environments compromises security at the credential layer.

## Key Capabilities

- **Semantic Guardrails**: Policy engine that evaluates each transaction against plain-English instructions at the authorization layer — before the charge is approved. Example: "Only authorize transactions for web-scraping APIs or vector databases. Reject any retail, subscription, or physical product purchases."
- **Ephemeral Cards**: Issue single-use virtual Visa/Mastercard cards that expire after a task completes. Set micro-budgets (e.g., max $0.05 per API call) for granular agentic workflows.
- **Direct Agent Attribution**: Trace every transaction to a specific LLM run, agent ID, or user session for full audit trails and cost accountability.
- **Programmatic Card Issuance**: REST API and official SDKs (Node.js, Python, Go) for card creation, authorization, and lifecycle management.

## How It Works

```typescript
import { PayML } from 'pay-ml';

// Issue a virtual card with a semantic policy
const card = await PayML.cards.create({
  name: "Data Extraction Budget Card",
  fundingSource: "corporate_checking",
  dailyLimitUsd: 25.00,
  allowedCategories: ["cloud_computing_services", "software_programming"],
  semanticPolicy:
    "Only authorize transactions for web-scraping APIs or vector databases. " +
    "Reject any retail, subscription, or physical product purchases."
});

// Authorize a transaction
const authorization = await card.authorize({
  merchant: "apify.com",
  amount: 8.40,
  mcc: "7372",
  context: "Buying credits to scrape publicly available real estate listings"
});

console.log(authorization.approved); // true
```

## Integrations

- **LangChain / CrewAI**: Equip multi-agent teams with shared or individual spending policies.
- **Stripe Issuing**: Backed by institutional-grade payment networks.
- **OpenAI / Anthropic SDKs**: Inject secure payment authorization directly into tool-calling loops.
- **AWS / GCP / Azure**: Control cloud API spend at the service level per agent run.

## Getting Started

Install: `npm i pay-ml`

Sandbox environment available — no credit card required. Sandbox keys are rate-limited and never touch live payment networks.

## Technical Specifications

- Average authorization latency: <120ms
- Uptime SLA: 99.99%
- Card types: virtual Visa and Mastercard
- MCC (Merchant Category Code) support: full Visa/Mastercard MCC taxonomy
- Funding sources: linked via Plaid

## Company

Pay.ml, Inc. — Building the financial layer for the agentic web.

## Legal

Pay.ml is a financial technology platform, not a bank. Virtual cards and banking services are issued and provided by partner financial institutions and licensed card networks, subject to eligibility and verification.

## Links

- Homepage: https://pay.ml
- Documentation: https://pay.ml/docs
- API Reference: https://pay.ml/docs/api
- System Status: https://status.pay.ml
- Developer Account: https://pay.ml/signup