> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usemandate.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Payment policy and mandate infrastructure for AI agents

Mandate is the trust layer for AI agent commerce. Give your AI agents verifiable identities, scoped spending authority, and a complete audit trail — without exposing your credentials.

## How it works

<Steps>
  <Step title="Create an agent">
    Register your AI agent with Mandate. Each agent gets a unique identity that sellers can verify before accepting payment.
  </Step>

  <Step title="Issue a mandate">
    Define exactly what the agent is allowed to spend: total budget, per-transaction cap, allowed sellers, categories, and expiry date.
  </Step>

  <Step title="Agent pays">
    When the agent makes a purchase, Mandate verifies its identity and enforces the policy in under 300ms before any money moves.
  </Step>

  <Step title="Audit everything">
    Every transaction is logged with full context — who authorized it, what policy applied, and the outcome.
  </Step>
</Steps>

## Core concepts

<CardGroup cols={2}>
  <Card title="Agents" icon="robot" href="/api/agents">
    AI entities that act on behalf of your application. Each agent has a unique identity and can hold multiple mandates.
  </Card>

  <Card title="Mandates" icon="file-contract" href="/api/mandates">
    Scoped spending authorities granted to agents. Set budgets, allowed sellers, categories, and expiry.
  </Card>

  <Card title="Policy Engine" icon="shield-check" href="/api/policy">
    Evaluates every transaction against the agent's mandate before it goes through. Returns approved or denied with a reason code.
  </Card>

  <Card title="Verification" icon="circle-check" href="/api/agents">
    Single endpoint for sellers to verify an agent's identity and charge in one call — no x402 implementation required.
  </Card>
</CardGroup>

## Sandbox vs production

All keys starting with `ky_sand_` are sandbox keys — they run the full policy engine but no real payments are settled. Use them freely for development and testing. Switch to a `ky_prod_` key when you're ready to go live.