AgentsPay

Explainer

AP2 vs ACP vs x402 and How the Three Agent Payment Protocols Compare

AP2, ACP, and x402 are the three standards shaping how AI agents pay. They solve different layers of the problem, and most real systems will touch more than one.

Agent Payments Console

Pick an agent

Payment intent

intent: ▌

Policy evaluation

Human approval required

This spend is over your approval threshold. Approve it to issue a scoped card, or deny it.

Scoped virtual card issued

AgentsPay

single-use

Wallet budget

spent of

Audit trail

In short

An agent payments protocol is a standard for how an AI agent authorizes and completes a payment. AP2 (Agent Payments Protocol), introduced by Google, uses verifiable mandates to prove an agent had user authorization for a purchase. ACP (Agentic Commerce Protocol), from OpenAI and Stripe, standardizes agent-driven checkout with merchants. x402 revives the HTTP 402 status code for instant crypto micropayments between machines. They address authorization, checkout, and settlement respectively, so they are complementary rather than mutually exclusive.

AP2 adds verifiable mandates for authorization

AP2, the Agent Payments Protocol led by Google, focuses on proving that an agent was authorized to spend. It uses cryptographically signed mandates that capture user intent and the limits of what the agent may buy, so a merchant or payment provider can verify the agent acted within a real, attributable instruction. AP2 is about trust and accountability at the authorization layer, and it is designed to be payment-method agnostic. For the full specification, the current mandate model and the FIDO Alliance handover, see our guide to Google AP2.

ACP standardizes agentic checkout

The Agentic Commerce Protocol, developed by OpenAI and Stripe, standardizes how an agent completes a purchase with a merchant. It defines how product information, carts, and checkout flow between an AI surface and a seller so an agent can buy without a custom integration per store. ACP operates at the checkout and commerce layer, making merchant transactions interoperable for agents.

x402 brings HTTP 402 machine micropayments

In practice most agents reach x402 through a tool call, so it is worth reading this alongside how MCP payments work. x402, introduced by Coinbase and now governed by the Linux Foundation, revives the long-reserved HTTP 402 Payment Required status code to let machines pay for resources inline over HTTP. A server can respond with a 402 and a price, and the client settles instantly, typically with stablecoins, before the request proceeds. x402 targets the settlement layer for low-value, high-frequency agent and API payments. For the response format, the facilitator role and what changed in v2, see our full guide to the x402 protocol.

Where the card networks fit

These protocols are open standards, but the money still settles over a payment rail, and the two biggest card networks now have their own agentic programs that implement them. Visa Intelligent Commerce supports payments initiated through several of these protocols and issues scoped, tokenized card credentials to agents. Mastercard Agent Pay does the same with Mastercard Agentic Tokens and adds agent-to-agent settlement across cards, bank accounts and stablecoins. If you are weighing which rail to build on rather than which protocol, our comparison of Visa Intelligent Commerce vs Mastercard Agent Pay breaks down how they differ and what neither one governs. Both networks also joined the launch coalition for the Universal Commerce Protocol, the Google and Shopify standard announced in January 2026 that wraps catalog, cart, checkout, and orders into one specification and is documented as interoperable with AP2. That is the layer above the three here, and it is the one most US merchants will be asked about next.

How AgentsPay fits across all three

These standards define how agents authorize, check out, and settle, but none of them decides whether a given spend should happen. AgentsPay is the rail-neutral control plane that sits across them: it enforces policy, applies hard spend limits, pauses for human approval, and writes an immutable audit trail regardless of which protocol moves the money. You adopt AP2, ACP, or x402 as they fit, and keep one place that never moves money without policy.

Whatever standard moves the money, AgentsPay is the rail-neutral control plane that keeps it governed. See how it works and the control surfaces that enforce policy, approvals, and audit on every transaction.

Frequently asked

Questions people ask about AP2 vs ACP vs x402

What is the difference between AP2 and ACP?

AP2 and ACP solve different layers. AP2, led by Google, proves authorization: a cryptographically signed mandate shows an agent had real user permission to buy within stated limits. ACP, maintained by OpenAI and Stripe, standardizes checkout between an agent and a merchant. A stack can use both, with AP2 proving the mandate and ACP completing the purchase.

Is x402 a replacement for card payments?

No. x402 targets inline machine settlement for small, high-frequency payments such as API calls and data access, usually in stablecoins over HTTP. Card rails still win for business purchases because charges are disputable, settlement is delayed and chargeback rights apply. Most real systems use x402 for micropayments and cards for everything a finance team has to reconcile.

Which agent payments protocol should I adopt?

Pick by the problem you have. If you sell to agents, implement ACP so agents can check out with you, and see our guide to the Agentic Commerce Protocol. If you need to prove an agent was authorized, look at AP2. If your agents pay per API call, look at x402. If your agents are the ones spending, none of them controls the budget and you need a policy layer on top.

Keep reading

More explainers

Plaid MCP Server

Plaid MCP Server

Plaid now offers a hosted MCP server for Claude, ChatGPT and agents built on the OpenAI and Anthropic APIs, plus a local MCP server for development. Engineering teams connecting them ask a fair question before anything ships: can an agent move money through Plaid, and if it can, what stops it? We read both servers' published tools, probed the hosted endpoint, and measured Plaid's public OpenAPI specification to answer it.

Read

Airwallex MCP Server

Airwallex MCP Server

Airwallex now ships AgentOS, a production MCP server, a CLI and a set of agent skills that let Claude, Cursor or Codex work directly on your Airwallex account. Before a finance team connects one, someone has to answer the question every controller asks in the review: what can this agent actually spend, and who decides the limit? We read the OAuth metadata of both Airwallex MCP servers and the card control documentation to find out.

Read

Tipalti MCP Server

Tipalti MCP Server

Tipalti now documents an official MCP server that connects Claude to your payables data, and its own AI agents already approve bills and match POs inside the product. Before an AP team lets any agent near supplier payments, the controller has to answer one question in the review: what stops it paying? We read Tipalti's MCP documentation, its OAuth metadata and every published REST operation to find out.

Read

Mercury MCP Server

Mercury MCP Server

Mercury now ships a hosted MCP server that connects Claude, ChatGPT, Cursor and a dozen other AI tools to your business bank account, and since August it issues agent cards an AI can check out with on its own. Before you hand either one to an agent, a founder or controller wants one answer: what actually stops it spending? We read Mercury's MCP documentation, its OAuth metadata and every operation in its public API on 15 October 2026 to find out.

Read

Bill.com MCP Server

BILL MCP Server

BILL now ships an official MCP server that connects your payables, receivables and card spend to Claude and ChatGPT, and third-party MCP servers already add write actions such as Create Payment on top of the BILL API. Before an AI agent touches the system that pays your vendors, a controller wants one answer: what stops it spending? We read every operation BILL publishes for its v3 API to find out.

Read

Deel MCP Server

Deel MCP Server

Deel now runs an official MCP server, so Claude, ChatGPT or Cursor can read contracts, list people, pull payslips and, with the right grant, create payments to your contractors and employees. For a finance lead or an HR operations owner the question before connecting it is simple: what stops the agent from paying the wrong person, or the right person too much? We read Deel's MCP documentation, called the live server, and read its OAuth metadata and the payment endpoints behind the tools.

Read

Rippling MCP Server

Rippling MCP Server

Rippling runs a first-party MCP server, so Claude, ChatGPT or Codex can look up people, check leave balances, file time off, draft a hire and edit custom objects as the employee who connected it. For an HR or IT lead the questions before turning it on are practical: what can the agent actually change, who decides, and where does money enter the picture. We read Rippling's MCP documentation end to end and called the live server with controls.

Read

Zapier MCP and Zapier Agents

Zapier MCP

Zapier MCP plugs Claude, ChatGPT, Cursor or any MCP client into more than 9,000 apps, and Zapier Agents run the same actions on a schedule. For finance apps that means an AI can create a Stripe payment, refund a Square charge, move money between Xero bank accounts or issue a Brex card. We read the action catalog Zapier publishes for nine finance apps on 13 October 2026 to find where a dollar limit lives.

Read

Sierra AI agents

Sierra AI

Sierra builds customer service AI agents for large consumer brands, and those agents do more than answer questions. They cancel orders, process returns, change bookings and hand out travel credits, and Sierra bills per outcome when they do. We read the only public code Sierra publishes about how such an agent is governed, its tau2-bench benchmark, on 12 October 2026, to see where the dollar limit lives.

Read

Workato MCP and Genies

Workato MCP and Workato Genies

Workato now sells itself as the control and execution plane for enterprise AI agents: Genies built in Agent Studio, more than 80 prebuilt MCP servers, and an MCP Gateway in front of all of them. Several of those servers can refund a Stripe charge, authorize a Xero bill or submit a Coupa requisition. We read Workato's own documentation on 9 October 2026 to find where a dollar limit lives.

Read

IBM watsonx Orchestrate

IBM watsonx Orchestrate

IBM sells watsonx Orchestrate as the control plane for every AI agent in the enterprise, and its agents now call tools, run flows and talk to other agents over MCP and A2A. We downloaded the developer kit IBM published on 7 October 2026 and read every agent and flow model in it, looking for one thing: where a dollar limit would go.

Read

Amex Agentic Commerce

American Express Agentic Commerce

American Express launched its Agentic Commerce Experiences (ACE) developer kit and Amex Agent Purchase Protection on 14 April 2026. It is the first card network to write "how much can be spent" into the agent's purchase intent and to promise a credit when a registered agent buys the wrong thing. We checked every public Amex host for the specification, an MCP server and discovery files on 7 October 2026. Here is what is actually live, and what a business still has to add before it lets agents spend.

Read

Concur AI and Joule

Concur AI and Concur Joule

SAP Concur is putting Joule agents into the expense report itself this year: one builds the report, one audits receipts before submission, and an AI approval assistant reaches US approvers in Q4 2026. Outside agents connect through Microsoft 365 Copilot or the Concur API. We read every OAuth scope SAP publishes for Concur to see where an agent's dollar limit would live. It is not in the scopes and it is not in the approve call.

Read

Coupa AI and Navi

Coupa AI and Coupa Navi

Coupa's Navi agents moved from answering questions to running payment batches this year, and since the September 2026 release Microsoft Copilot and custom agents can read and write Coupa data through MCP. We probed the Coupa MCP authorization servers on six live US tenants to see where the dollar limit sits for an outside agent. It is not in the token.

Read

Navan and Expensify

Navan AI and Expensify

Navan and Expensify both opened their travel and expense data to AI assistants this summer, and both are moving from answering questions to acting on reports. Expensify Agents can already submit, approve, reject and route expense reports. Navan says write tools for approving expenses and booking travel are next. We measured where the dollar limit sits for those agents, and in the agent itself it does not sit anywhere.

Read

HubSpot Breeze

HubSpot Breeze

HubSpot Breeze agents now resolve tickets, recommend leads and, through agent tools and the HubSpot MCP server, write carts, orders, quotes and line items. HubSpot meters what the agents cost you in HubSpot Credits, with an account cap and per-feature caps. It does not meter what they spend or commit on your behalf, and that gap is what this page measures.

Read

Intercom Fin

Intercom Fin

Intercom Fin no longer just answers questions. Through Fin Procedures and data connectors it looks up a Stripe subscription, cancels it and issues the refund while the customer is still typing. Intercom charges you 99 cents for that outcome. The refund itself can be any amount, and that is the number worth putting a limit on.

Read

Zendesk AI Agents

Zendesk AI agents

Zendesk AI agents now do more than answer. With AI Agents Advanced they run generative procedures that call your APIs, cancel Shopify orders and issue refunds while the customer is still in the chat. Zendesk bills you per automated resolution. Nobody bills you for the refund, and that is the number worth controlling.

Read

Brex AI Agents

Brex AI agents

Brex made the opposite bet to Ramp. Its MCP server lets Claude, ChatGPT or Cursor read almost everything in your Brex account and change almost nothing that matters for money. The Brex API is a different story: the same company publishes endpoints that create spend limits, issue cards and send wires. Which of those two doors your AI agent walks through decides what it can spend.

Read

Ramp AI Agents

Ramp AI agents

Ramp has built the most complete spend model for AI agents of any finance platform we have measured: agent identities, agent roles, agent cards, and funds with real dollar limits. That is exactly why the permission question matters here more than anywhere else. The limits are real, so the question is who can change them, and on Ramp the answer can include the agent.

Read

SAP Joule

SAP Joule

SAP Joule is moving from answering questions to acting inside S/4HANA, Ariba and Concur, which means SAP AI agents now sit one function call away from purchase orders and supplier payments. SAP publishes the developer tooling for Joule Studio on npm, so we downloaded it and asked the question a controller asks before an agent goes live: where does the dollar limit go? The answer is that SAP models money in great detail one layer down, and the Joule layer on top carries none of it.

Read

UiPath Agentic Automation

UiPath Agentic Automation

UiPath ships more agent governance than almost any vendor we have measured. It has a policy engine, four enforcement actions including human approval, and consumption metering accurate to a fraction of a unit. So we downloaded both SDKs UiPath publishes and asked the one question a finance lead asks before an agent goes live: can any of it be given a number in dollars? The answer is specific, and it is not the answer the word "budget" suggests.

Read

Workday AI Agents

Workday AI Agents

Workday now registers AI agents the way it registers employees, in the Agent System of Record, and more than 65 partners are wiring their agents into it. We downloaded the API specification Workday publishes for that registry to answer the question a CFO asks before any agent goes live: where is the spending limit? Workday's documents carry plenty of money controls. The agent record carries none.

Read

Oracle AI Agent Studio

Oracle AI Agent Studio

Oracle AI Agent Studio is included with every Oracle Fusion Cloud subscription, and it now lets a finance team build agents that call any REST API or MCP server. We read the API specification Oracle publishes for those agents to answer the question a controller asks before switching one on: what stops an agent spending money? Inside Fusion, quite a lot. Outside Fusion, nothing in the agent model does.

Read

ServiceNow AI Control Tower

ServiceNow AI Control Tower

ServiceNow AI Control Tower is the most complete agent inventory and risk console a large US enterprise can buy, and it now reaches across AWS, Google Cloud and Azure. We read the schema ServiceNow ships to developers to answer the one question the rollout meeting always ends on: can it stop an agent from spending money? It cannot, and the reason is written into the data model.

Read

Gemini Enterprise

Gemini Enterprise

Google did something in August 2026 that the other agent platforms have not done: it shipped a hard monthly spend cap that genuinely stops usage instead of emailing you about it. That deserves credit, and it also moves the interesting question one step along. A cap that stops something is only as useful as the thing it is scoped to, so we went and measured what Google can actually point that cap at, in the API model Google publishes for anyone to read.

Read

Salesforce Agentforce

Salesforce Agentforce

Agentforce is the largest agent platform any US enterprise is likely to already own, and it moved to consumption billing, which means the meter now runs on what your agents do rather than on how many seats you bought. That raises a finance question the rollout deck rarely answers: when an Agentforce agent is loose in production, what actually stops it spending. We went and measured the answer in Salesforce own published object model rather than guessing at it.

Read

AWS AgentCore

AWS AgentCore

Amazon shipped the missing piece in August 2026. Bedrock AgentCore Payments went generally available, and it is a real payments product: an agent can now hold a wallet, meet an HTTP 402, pay, and carry on reasoning without a human in the loop. So the question a platform lead has to answer stopped being whether AWS gives agents money and became a narrower, more awkward one: how much of a spend policy did AWS actually ship? We went and measured it, property by property, in the API model AWS publishes.

Read

Microsoft Agent 365

Microsoft Agent 365

Microsoft shipped a control plane for AI agents, and it is a good one. It gives every agent an identity, a registry entry, an owner, a sponsor and a Conditional Access policy. Then somebody in finance asks the obvious follow-up question: fine, but what stops the agent from spending money? This page answers what Agent 365 costs, what it governs, and what we measured when we went looking for a dollar amount anywhere in Microsoft's agent governance surface.

Read

QuickBooks MCP Server

QuickBooks MCP server

Connecting an accounting system to an AI assistant is now a ten minute job. Deciding what that assistant is allowed to do once it is connected is the part nobody writes about, and it is the part your controller will ask about first. This page compares what the official QuickBooks, NetSuite and Xero MCP servers actually hand a model, measured rather than summarized from marketing pages.

Read

Payment MCP Servers

payment MCP servers

Every large payment company shipped an MCP server in the last eighteen months, and almost every write-up of them is a setup tutorial. The setup is the easy part. The question worth answering before you connect one to a production account is narrower and much less comfortable: what, exactly, can the model on the other end of that connection do to your money?

Read

PayPal Agentic Commerce

PayPal Agentic Commerce

PayPal made a bet that most merchants would rather not implement a commerce protocol at all. Where Stripe and OpenAI shipped a spec for you to build against, PayPal shipped two products that sit on top of the checkout you already have, and then bought a company to make the catalog half work. That choice is the whole story: it explains why Agent Ready needs almost no engineering from you, why there is nothing for an agent to discover about your store on the open web, and why the thing PayPal will not do for you is the thing that gets expensive later.

Read

Shopify Agentic Commerce

Shopify Agentic Commerce

Shopify switched agentic commerce on by default, so your store is probably already selling to AI assistants whether or not anyone on your team configured it. Instead of restating the announcement, we checked something you can check too: on September 2, 2026 we requested the machine-readable capability file that Shopify publishes for real storefronts, on fourteen well-known US brand domains, and read what it exposes to an agent. Eleven answered correctly. The three that did not share one trait, and it is quietly costing them agent traffic.

Read

Web Bot Auth

Web Bot Auth

Web Bot Auth is the reason your agent either gets served or gets throttled with the scrapers. Almost everything written about it repeats the same architecture diagram, so we did something different: on September 1, 2026 we fetched the published key directories of more than twenty major AI operators and infrastructure vendors to see who is genuinely signing their traffic. Four were. The results are in the first table.

Read

Tempo Blockchain

the Tempo blockchain

Tempo is the payments chain Stripe and Paradigm built, and it shipped with a protocol that lets software pay for things on its own. It settles machine payments in under a second. It has nothing at all to say about whether your agent should have paid.

Read

AI Agent Governance

AI agent governance

Every agentic AI governance framework published so far governs the same four things: identity, tools, data and prompts. Not one of them carries a budget. Here is what the real frameworks say, which guardrails actually bind at runtime, and what to do about the last mile none of them reach.

Read

A2A Protocol

A2A Protocol

Most explanations of the A2A protocol stop at the sentence that agents can now talk to each other. That was true in April 2025 and it is no longer the interesting part. A2A shipped version 1.0 in April 2026 under Linux Foundation governance, it runs in production inside Azure AI Foundry and Amazon Bedrock AgentCore, and the questions engineers actually get stuck on are narrower: what an Agent Card commits you to, when to reach for MCP instead, and what happens the first time one of your agents has to pay another one for the work. That last question has a specific answer, and it is not in the core spec.

Read

Mastercard Agent Pay

Mastercard Agent Pay

Nearly every article about Mastercard Agent Pay is a retelling of one press release from April 2025, the one where Mastercard said AI agents would be able to shop with Agentic Tokens and named Microsoft as the first platform. That was sixteen months ago, and four more things have shipped since. Reading only the launch coverage leaves you with roughly a quarter of the picture, and the missing three quarters are the parts that decide whether you can actually put this into production.

Read

Visa Intelligent Commerce

Visa Intelligent Commerce

Almost everything written about Visa Intelligent Commerce is a retelling of the April 2025 announcement, when Visa said AI agents would be able to pay with a Visa credential. Three more things have shipped since, including an open agent-identity protocol built with Cloudflare that most coverage does not mention at all. This page is the current version, checked against Visa’s own developer documentation and newsroom in August 2026.

Read

Stripe agentic commerce

Stripe agentic commerce

Most writing about Stripe and agentic commerce is still a retelling of the September 2025 launch week, when Stripe and OpenAI shipped Instant Checkout and published the Agentic Commerce Protocol together. Stripe has built a good deal more since then, and some of it points in a direction the launch coverage never anticipated. This page is the current version, checked against Stripe’s own documentation in August 2026.

Read

ChatGPT Instant Checkout

ChatGPT Instant Checkout

Almost every guide to ChatGPT Instant Checkout still reads like it was written the week it launched, walking merchants through how to apply and what the fee will be. OpenAI changed course in March 2026. Here is the accurate version: what Instant Checkout was, what the numbers actually looked like, what replaced it, and which parts of the stack are still very much alive.

Read

Google AP2

Google AP2

Most guides to Google AP2 still describe an Intent Mandate and a Cart Mandate, because most of them are rewrites of the September 2025 launch post. The specification moved. Here is what the Agent Payments Protocol actually defines today, and the one question it deliberately does not answer.

Read

Human in the loop AI

Human in the Loop AI

Every guide to human in the loop AI describes the same shape: the agent pauses, a person decides, the agent continues. The shape is right. What almost none of them ask is a harder question, which is where the pause is enforced, because a pause written into the agent's own code is a pause the agent is trusted to honor.

Read

AI agent cost

AI Agent Cost

Every cost guide for AI agents answers the same two questions: what does it cost to build, and what does it cost to run. Both are answerable, and both are on somebody's invoice. The third question is the one that ends up in a variance report, because the agent also spends your money, and nobody sends you a bill for that.

Read

Agentic checkout

Agentic Checkout

Nearly every guide to agentic checkout is written for the merchant who wants to receive these orders. Far fewer are written for the company whose agents are placing them, which is odd, because agentic checkout quietly removes the one screen where spending used to get a second look.

Read

API monetization

API Monetization

Most guides to API monetization argue about which pricing model wins. The harder question in 2026 is who is calling. An API priced for a signed-up developer with a key behaves very differently when the caller is an agent that showed up once, wants one record, and has no account.

Read

x402 protocol

x402 Protocol

x402 took the one HTTP status code the web never used and turned it into a payment rail machines can drive. The protocol is elegant and genuinely small. The part it deliberately leaves to you is the budget.

Read

AI procurement agents

AI Procurement Agents

Every major procurement suite shipped agents during 2026. Almost none of them answer the question your controller will ask first, which is what happens when the agent is wrong about a purchase and the money has already moved.

Read

Agentic payments

Agentic Payments

Agentic payments move money with no human at the checkout. The rails to do it all shipped during 2026. The part most teams have not solved is deciding, before the money moves, whether the agent was allowed to spend it.

Read

AI agent monetization

AI Agent Monetization

Every AI agent company is rewriting its price list. The models that survive are metered. The ones that quietly fail are the ones where nobody measured what a single task costs to serve.

Read

Agent payment platforms

AI Agent Payment Platforms

Five different kinds of product now call themselves an AI agent payment platform, and they solve five different problems. Picking the wrong category is the expensive mistake, not picking the wrong vendor inside a category.

Read

Universal Commerce Protocol

the Universal Commerce Protocol (UCP)

Google and Shopify shipped UCP as an open standard so an AI agent can check out at any merchant that supports it. Here is what the specification actually defines, where it is live for US buyers, and the one thing it deliberately leaves to you.

Read

MCP Payments

MCP Payments

MCP payments are how an AI agent discovers a payment tool and calls it to move money. The catch: the Model Context Protocol carries the tool call, not the spending decision, so nothing in the stack asks whether the purchase should have happened.

Read

Visa Intelligent Commerce vs Mastercard Agent Pay

Visa Intelligent Commerce vs Mastercard Agent Pay

Visa Intelligent Commerce and Mastercard Agent Pay are the two big card networks racing to let AI agents pay. They take different routes to the same idea, and neither one decides whether a given purchase should have happened.

Read

Agentic Commerce Protocol

the Agentic Commerce Protocol

ACP is the open standard behind agentic checkout in ChatGPT. It tells a merchant how to sell to an AI agent. It says nothing about whether your agent should have made the purchase.

Read

Machine payments protocol

Machine payments protocol

As software starts paying software, machine payments protocols define how value moves without a human at the keyboard. The harder question is how to keep that spending governed.

Read

Know Your Agent (KYA)

Know Your Agent

KYA, or Know Your Agent, extends the idea of customer due diligence to autonomous software. When an agent spends, you need to know which agent, on whose authority, and under what limits.

Read

Keep agent spending governed

Add policy, hard limits, human approval, and an immutable audit trail across any protocol or rail. Start in the sandbox today.

Never moves money without policy