PinkWallet Website Logo
Join early access
Last updated: 2026-09-28

Learn: Agentic payments, explained

Plain-language definitions and evidence-based guides to the protocols, wallets, and controls behind AI agents that can pay — x402, AP2, ACP, Visa TAP, Mastercard Agent Pay, MCP payments, and more.

Glossary

Definitions of the core terms in agentic payments.

Glossary

What Are Agentic AI Payments? Definition, Protocols, and How It Works

Agentic payments let AI agents initiate and execute transactions on a human's behalf. Definitions, protocols (x402, AP2, ACP), and open questions.

Glossary

What Is x402? The HTTP 402 Payment Protocol Explained

x402 is Coinbase's open protocol that uses the HTTP 402 status code to let servers charge AI agents and apps in stablecoins. Origins, mechanics, governance.

Glossary

What Is AP2 (Agent Payments Protocol)? Google's Agentic Payment Standard

AP2 is Google's open protocol for AI agents to make purchases using signed 'mandates.' Origins, mechanism, governance, and how it differs from x402/ACP.

Glossary

What Is the Agentic Commerce Protocol (ACP)? OpenAI + Stripe's Checkout Standard

ACP is OpenAI and Stripe's open standard letting AI agents complete purchases inside a chat interface. Origins, checkout flow, license, and comparisons.

Glossary

What Is an AI Agent Wallet? Definition and How It Works

An AI agent wallet is a crypto or payment wallet built for autonomous AI agents rather than humans, with spend limits and delegated signing. Examples inside.

Glossary

What Are AI Agent Spending Controls? Caps, Approvals, and Mandates Explained

AI agent spending controls are the rules — caps, allowlists, approval thresholds, velocity limits — that bound what an autonomous agent can pay for. How AP2, card networks, and wallets each implement them.

Glossary

What Is Know Your Agent (KYA)? AI Agent Identity Verification Explained

Know Your Agent (KYA) verifies which human or organization an AI agent represents before letting it transact. How Skyfire's KYA works and how it relates to KYC.

Glossary

What Is Mastercard Agent Pay? Agentic Tokens Explained

Mastercard Agent Pay lets AI agents transact using Agentic Tokens — permissioned, tokenized payment credentials. Announcement, mechanism, and named partners.

Glossary

What Is the Visa Trusted Agent Protocol? Definition and How It Works

Visa Trusted Agent Protocol lets merchants cryptographically verify legitimate AI shopping agents instead of treating them as bots. Mechanism, partners, and specs.

Glossary

What Are Payments Over MCP? How AI Agents Pay Through the Model Context Protocol

MCP payments let an AI agent call payment tools (create an invoice, charge a card, issue a refund) through the Model Context Protocol. Official servers, risks, and safeguards.

Guides

Longer, comparative pieces for developers and technical decision-makers.

Guide

Payment MCP Servers Compared (2026): Which Providers Let AI Agents Move Money

16 payment companies checked for an official MCP server that moves money, plus what auth and spending controls each one actually documents.

Guide

Stripe MCP Agent Keys: What Changes on October 31, 2026 and How to Migrate

Stripe MCP stops accepting full-access secret keys and non-Agent restricted keys on Oct 31, 2026. Here's who's affected, how to migrate, and how to scope agent spend.

Guide

x402 vs AP2 vs ACP (and MPP, UCP): Which AI Agent Payment Protocol Should You Use? (2026)

x402, AP2 and ACP solve different layers of AI agent payments: pay-per-request stablecoin settlement, proof of user authorization, and in-chat checkout. Side-by-side comparison, adoption data for 13 providers, and when to use which.

Guide

Agentic Payments Market Map 2026: Who Builds What in AI Agent Payments

A layer-by-layer market map of agentic AI payments in 2026 — protocols, card-network programs, agent wallets, payment MCP servers, and spend-control/policy tools — with 33 companies, primary sources, and status for each.

Guide

AI Agent Spending Policy: A Template for Finance Teams (Caps, Allowlists, Approvals, Audit)

A copy-paste policy template and 9 controls (budgets, allowlists, approval thresholds, credential scoping) for letting AI agents spendunder finance oversight.

Guide

Best Payment APIs for AI Agents (2026), by Use Case

There is no single best payment API for AI agents — the right pick depends on what the agent needs to do. A use-case-by-use-case guide to card checkout, stablecoin micropayments, agent wallets, MCP-based account access, and identity, sourced to each provider's own documentation.

Guide

How to Let an AI Agent Pay for API Usage Safely (x402, Prepaid Credits, Virtual Cards, MCP)

Five documented ways to let an AI agent pay for per-call or metered API usage without handing it an open-ended credential: x402 pay-per-request, provider spend limits, single-use virtual cards, an MCP policy layer, and human approval — compared, with a setup checklist.

Guide

Spending controls for AI agents: why agentic payments need a policy layer

Guide

How AI agents make payments through MCP

AI agents pay through MCP by calling a payment tool that budget and approval rules check first. Pink Agentic AI Payments adds that policy layer over any MCP payment server.

Guide

Agentic Payments Timeline (2024–2026)

A dated, source-linked timeline of protocol launches, official payment MCP servers, card-network agent programs and governance changes in AI agent payments.

Guide

x402 vs MPP: Which HTTP 402 Payment Protocol Should Your AI Agent Use? (2026)

x402 (Coinbase, now the x402 Foundation) and MPP (Stripe and Tempo) both turn HTTP 402 into a machine payment layer. Side-by-side comparison of settlement rails, SDKs, spend controls and governance, sourced from each protocol's own docs.

Guide

AI Agent Payment Incident Runbook: What to Do When an Agent Misbehaves With a Payment Credential

An operational runbook for developers and finance/IT teams: what to do in the first 15 minutes after an AI agent misuses a payment credential, how to investigate, what you can recover, and how to prevent a repeat.

Guide

AP2 Mandates and Budget Constraints: 4 Gotchas From the GitHub Issues (2026)

Four confusions developers keep hitting when implementing AP2 (Agent Payments Protocol) mandate and budget-constraint semantics — budget.max unit mismatches, a false-negative violations list, mandate replay, and a skippable checkout-binding check — documented against the AP2 reference implementation and filed as public GitHub issues.

Guide

Virtual Cards for AI Agents Compared (2026): Which Card Issuers Let an Agent Spend Under Limits

A sourced, use-case-by-use-case comparison of virtual/programmable card providers that let an AI agent spend within limits — Privacy.com, Ramp, Brex, Stripe Issuing, Lithic, Marqeta, Crossmint, and AgentCard — with controls, signup path, and MCP/API access for each, sourced to each provider's own docs.

Guide

AI Agent Spend Governance Tools Compared (2026): Policy Engines and Control Layers for Agent Payments

A sourced comparison of 11 providers, including Pink Agentic AI Payments, that document spending governance for AI agents — Payman, Skyfire, Coinbase CDP, Circle, Crossmint, Lithic, Stripe Issuing, AgentCard, Tempo, Openfort, and Nevermined — by control layer (card network, wallet/onchain, API policy, MCP), with the exact caps, approvals, and audit features each vendor documents.

Guide

How to Set Per-Agent Spending Limits for an AI Agent (MCP, Cards, Wallets): Field by Field

The exact field or flag each provider documents for capping what an AI agent can spend — Stripe, Privacy.com, Lithic, AgentCard, Crossmint, Coinbase CDP, Circle, Tempo, Payman, Skyfire, AP2, and x402 — with its unit, its period options, and where it's actually enforced.

Guide

Who Sets the Standards for AI Agent Payments? The Bodies, the Specs, the Status (2026)

A sourced map of who actually governs each AI-agent payment spec in 2026 — x402, AP2, ACP, MPP, UCP, MCP, Visa's Trusted Agent Protocol and Mastercard Agent Pay — with the steward, licence, status and the last governance change for each.

Guide

How to Let AI Agents Pay Vendors Automatically: A Setup Guide for Companies (2026)

A five-step, vendor-neutral setup guide for finance, IT, and procurement leads who want an AI agent to pay vendors automatically — which rail to use per vendor type, how to write the spending policy, how to wire approvals, how to audit, and how to plan for incidents. Every control cited is sourced to the provider's own docs.

Guide

Stablecoin Payment Providers for Machine-to-Machine and AI Agent Payments (2026): Who Supports What

A sourced shortlist of 8 providers that document stablecoin support for machine-to-machine or AI agent payments — Coinbase, Circle, Stripe, Skyfire, Crossmint, Nevermined, Openfort, and Tempo — compared by stablecoin/chain, M2M mechanism, key custody, per-agent spending controls, and availability, each claim linked to the vendor's own docs.

Guide

Best Payment Setup for Autonomous AI Procurement (2026): Providers Compared for Companies

A use-case-led, sourced comparison of 10 payment providers — Stripe Issuing, Lithic, Privacy.com, AgentCard, Crossmint, Coinbase CDP, Circle, Payman, Skyfire, and Openfort — for companies letting an AI agent buy SaaS, API usage, and larger purchases on their behalf. Every control is sourced to the provider's own docs.

Guide

Pink vs Ramp for AI Agent Spend: Which One Should Control Agent Payments?

Ramp Agent Cards issue real single-use corporate cards to AI agents. Pink enforces spending rules before a credential exists. Sourced comparison.

Guide

Pink vs Brex for AI Agent Spend: Can Brex's MCP Server Let an Agent Pay?

Brex's MCP server gives agents read and expense-tracking access today, not payment execution. Pink checks rules before issuing a credential. Sourced.

Guide

Pink vs Coupa Navi MCP for AI Agent Spend Controls

Coupa's Navi Connect MCP exposes 30+ procurement tools to AI agents, inheriting the signed-in user's permissions. Pink checks rules first. Sourced.

Guide

Pink vs Stripe Agent Toolkit for AI Agent Payments

Stripe's Agent Toolkit and MCP server use a human-confirmation gate on sensitive writes. Pink checks rules before a credential exists. Sourced.

Guide

Pink vs AWS Bedrock AgentCore Payments for AI Agent Spend

AWS AgentCore Payments moves stablecoin micropayments at machine speed, GA since Aug 2026. Pink adds a human-approval rule layer. Sourced comparison.

Guide

Pink vs Marqeta MCP Server for AI Agent Card Issuing

Marqeta's MCP server issues virtual cards with velocity and spend controls, in beta. Pink checks rules before any credential exists. Sourced.

Guide

How to Let AI Agents Pay Vendors Automatically With Pink's Spending Limits

A direct, step-by-step setup: connect an AI agent to Pink over MCP or REST, set per-agent budgets and payee rules, and see real sandbox outputs for allowed, pending-human, and blocked payments.

Guide

A Policy Engine for AI Agent Payments: How Pink's Rules Layer Works

How Pink's server-side policy engine evaluates every AI agent payment, circuit breakers, amount tiers, payee rules, fraud signals and time windows, with a real dry-run trace and live sandbox outputs.

Guide

Pink: An MCP Server That Lets AI Agents Like Claude Pay Invoices Within Spending Rules

How Pink's MCP server lets an AI agent in Claude, ChatGPT or another MCP client check and pay an invoice within budgets, caps and approval rules, with the 7 MCP tools and real connection steps.

Guide

How to Give an AI Agent Its Own Wallet and Spending Limits with Pink

How to give an AI agent its own agent key, vault and budget in Pink: real field names, a worked procurement-agent config, and concurrent-load tests including a shared ceiling across multiple agents.

Guide

Risks and Controls for Agent-Initiated Payments: A Checklist from Pink

The 6 recurring risks in agent-initiated payments (budget runaway, payee drift, prompt injection, missing audit trail, no human in the loop, over-scoped credentials) mapped to Pink's server-side controls, with concurrent-load test results and an honest list of what isn't covered yet.

Use cases

How companies in different industries use Pink Agentic AI Payments — see the full Use cases hub.

Use case

AI Agent Payments for Corporate Travel Booking

How a company lets a travel-booking agent pay for flights and hotels inside per-trip caps, currency rules and manager approval above a threshold.

Use case

AI Agent Payments for E-Commerce

How an e-commerce company lets ads, creator, procurement and returns agents pay against revenue, signed briefs and POs, with tiered refunds and carrier statement checks.

Use case

AI Agent Payments for Logistics and Freight

How a logistics or freight company lets agents pay carrier invoices, fuel and toll spend and multi-currency payees, with statement matching and after-hours human checks.

Use case

AI Agent Payments for Manufacturing Procurement

How a manufacturer lets purchasing agents pay suppliers against a PO and warehouse scan, from an allowlist, with fraud checks and 2-of-3 approval above a threshold.

Use case

AI Agent Payments for Marketing Agencies

How an agency lets ad-buying and payout agents spend per client inside separate vaults and monthly caps, with the client's approver in the loop.

Use case

AI Agent Payments for Multi-Location Retail

How a multi-store retailer gives each store its own agent, vault and budget, with warehouse-scan checks on reorders and regional manager sign-off above a threshold.

Use case

AI Agent Payments for Professional Services Firms

How a law, accounting, or consulting firm lets agents pay SaaS bills, research purchases and contractors, with duplicate-invoice checks and partner sign-off.

Use case

AI Agent Payments for Property Management

How a property manager lets a maintenance agent pay allowlisted vendors, catches bank-detail-change fraud, and keeps a reserve vault agents can't touch.

Use case

AI Agent Payments for Restaurants and Cafés

How a coffee shop or restaurant lets purchasing, marketing and payroll agents pay suppliers and bills inside rules a human set, with approvals for anything unusual.

Use case

AI Agent Payments for SaaS Startups

How a SaaS startup lets engineering, ops and recruiting agents pay for cloud spend, subscriptions and contractors, with a $200/day rule that stopped a $49,600 bug.