Agentspay

Explainer

AWS AgentCore Spend Controls: Amazon Bedrock AgentCore Gateway, Identity and Policy Limits

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.

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

AWS AgentCore is Amazon’s managed runtime and control plane for AI agents, billed per component with no subscription and no minimum. Its money component, Bedrock AgentCore Payments, became generally available on 18 August 2026 and supports two protocols, x402 and the Machine Payment Protocol. We read the published AgentCore API model on 9 September 2026 and measured its entire spend-policy surface. It is one property: SessionLimits.maxSpendAmount, a value plus a currency, and the Currency enum contains exactly one member, USD. That limit is optional on CreatePaymentSession, so a session can be opened with no ceiling at all. A session must expire, between 15 and 480 minutes, and every new session starts with a fresh maximum, because no property anywhere in the model expresses a cumulative, daily or monthly total. There is no merchant field, no category or MCC field, no counterparty, and no approval threshold on the payment path. The instrument enum has one member, EMBEDDED_CRYPTO_WALLET, and the balance token enum has one member, USDC, so no card and no bank rail is reachable through it. AgentCore governs where an agent may connect and, now, that it may pay. It still does not govern what it is allowed to buy.

What AWS AgentCore is, and which parts touch money

AWS AgentCore is the set of managed services Amazon put underneath agents so teams stop hand-rolling the same infrastructure. There are thirteen billable capabilities and they divide cleanly by what they govern. Runtime, Browser and Code Interpreter govern execution: they give the agent somewhere isolated to run, browse and execute code, billed on active vCPU-hours and memory GB-hours. Gateway, Identity and Policy govern access: which tools exist, which credentials the agent may borrow, which calls are authorized. Memory, Observability, Evaluations and Agent Registry govern state and oversight.

Exactly one of them governs money, and it is the newest. Amazon Bedrock AgentCore Payments reached general availability on 18 August 2026, having been in preview since 7 May. It gives the agent a stablecoin wallet through either Coinbase CDP or Stripe Privy, and when the agent hits a paid resource that answers HTTP 402, AgentCore negotiates the payment and hands back proof without breaking the reasoning loop.

That is a genuinely useful thing to have shipped, and the design is careful. It is also much narrower than the phrase "agent payments" suggests, and the narrowness is measurable rather than a matter of opinion. So we measured it.

How we measured the spend-policy surface

AWS publishes the machine-readable model for every service API in the botocore repository. That model is the contract: if a property is not in it, no console toggle, blog post or roadmap slide can make the API accept it. So the honest way to answer "what spend rules can AgentCore enforce" is to read the model and count.

On 9 September 2026 we pulled two files from the botocore develop branch: the AgentCore data plane (bedrock-agentcore, API version 2024-02-28, 480,002 bytes) and the AgentCore control plane (bedrock-agentcore-control, API version 2023-06-05, 994,087 bytes). Between them they describe 237 operations, 2,216 shapes and 4,162 member properties.

Then we ran two passes. The first looked for any money vocabulary at all: amount, currency, budget, spend, price, cost, payment, invoice, balance, wallet, fee, charge, billing, limit. The second looked for the other dimensions a real spend policy needs: merchant, category, MCC, vendor, counterparty, allow list, approval, approver, daily, monthly, cumulative, aggregate, recipient, payee, purpose.

The first pass hit 70 properties in the data plane and 48 in the control plane, across 11 payment operations. So unlike the Microsoft agent surface, AgentCore plainly does model money. The interesting result is the second pass, and how few properties carry the whole policy.

The result: one amount, one currency, one clock

The entire spend ceiling in AgentCore Payments is a structure called SessionLimits, and it has exactly one member: maxSpendAmount. That member is of type Amount, which is itself two fields, value and currency. The Currency enum contains one value, USD. The readout side mirrors it: AvailableLimits.availableSpendAmount tells you what is left.

Three details in that paragraph matter more than they look.

First, limits is optional on CreatePaymentSessionRequest. The required fields are the payment manager ARN and an expiry time. The ceiling is not required. A session created without it is a wallet handed to a model with no stated maximum, and the API will accept that call.

Second, the ceiling is per session, and a session is short-lived by construction: expiryTimeInMinutes is required and constrained to a minimum of 15 and a maximum of 480. Eight hours is the longest a payment session can live.

Third, and this is the part that changes how you should budget: our second regex pass found no cumulative, daily, monthly, weekly or aggregate property anywhere in either model. The one match in the data plane was InsightsFailureSignal.category, which is an observability field about failures, not a spend rule. So each new session starts fresh at the full maximum. An agent that runs continuously for a week does not spend one cap. It spends the cap as many times as it opens a session.

What the payment rails actually are

The enums settle a question a lot of buyers assume the other way. PaymentType has two members, CRYPTO_X402 and MPP. PaymentInstrumentType has one member, EMBEDDED_CRYPTO_WALLET. InstrumentBalanceToken has one member, USDC. The wallet networks are Ethereum and Solana, and the chain identifiers are Base, Base Sepolia, Ethereum, Solana and Solana Devnet.

Read together: AgentCore Payments is a stablecoin rail. There is no card, no ACH, no wire and no invoice in it. We checked that claim rather than inferring it, because it is the kind of statement that is easy to get wrong. The literal string "card" does appear fifteen times in the data-plane model, and every single occurrence is the A2A Agent Card, the metadata document that describes an agent to other agents. There is no creditCard, no cvv, no iban, no merchant and no mcc anywhere in the file.

Your agent can therefore pay for a metered API, an MCP server, a piece of web content or another agent. It cannot check out at a US merchant, pay a supplier invoice, or put anything on a corporate card, because none of those rails are reachable from this API. If your agent buying story involves a purchase order, a vendor or a card statement, AgentCore Payments is not the component that does it, and that is worth knowing before the architecture review rather than after.

Why the control-plane approval field is not the approval you want

Our dimension pass did find approval vocabulary, and it would be easy to misreport. ApprovalConfiguration.autoApproval exists in the control plane, and it appears on CreateRegistry, GetRegistry and UpdateRegistry. So AgentCore does have an approval concept.

It governs the Agent Registry: whether an agent published into a registry is auto-approved or has to be reviewed by a human. That is a supply-chain control over which agents your organization is willing to list and reuse. It is a good control and it is not a spend control. Nothing in the eleven payment operations consults it. ProcessPayment takes a payment manager ARN, a session ID, an instrument ID, a payment type and a payment input. There is no approver, no threshold and no pending state in that call: it either validates against the session ceiling or it does not.

The same care applies to AgentCore Identity and AgentCore Gateway, which are the two components teams most often assume already solve this. Identity vaults and brokers OAuth tokens and API keys so an agent can act against a third-party service without a human pasting a secret. Gateway turns your APIs and Lambda functions into MCP tools and applies rate limits to them. We found BatchPutGatewayRateLimits and CreateGatewayRateLimit in the control plane, so Gateway can cap how many times a tool is called. Neither component can cap what a call is worth. A rate limit on a procurement tool is not a budget, because ten calls that each spend 50,000 dollars pass a limit of a hundred calls per minute without complaint.

What this means for the AgentCore bill

There is a second, quieter consequence. AgentCore Payments is the only component on the AWS pricing page with pass-through pricing: AWS adds no charge of its own for CreateInstrument or ProcessPayment, and you pay whatever Coinbase CDP or Stripe Privy charges. Every other component has a published AWS unit rate. Runtime is 0.0895 dollars per vCPU-hour and 0.00945 dollars per GB-hour. Gateway is 0.005 dollars per 1,000 API invocations. Identity is 0.010 dollars per 1,000 token or API key requests, and free when used through Runtime or Gateway. Policy is 0.000025 dollars per authorization request.

So the money your agents spend does not appear on the AgentCore line of your AWS bill at all. It lands in a wallet provider’s ledger, denominated in USDC, on Base or Solana. Finance is going to reconcile a cloud invoice that shows compute and invocations, and separately try to explain a stablecoin ledger that shows the actual purchases. We wrote up the full component-by-component numbers in our AgentCore pricing breakdown, including the components most calculators leave out.

Where a spend policy layer fits on top

None of this is a criticism of the design. A session-scoped USD ceiling on a stablecoin wallet is a sensible primitive, and primitives are what infrastructure should ship. The gap is that a primitive is not a policy, and the distance between them is where the finance objection lives.

A policy answers questions AgentCore’s model has no field for. Which counterparties is this agent allowed to pay at all. What is its total for the month, not for the next eight hours. Which purchases need a named human to approve before they clear, and which clear silently. What happens on the tenth identical charge in a minute. Which cost center owns this, in the general ledger, in dollars.

That is what we build. Agentspay sits between the agent and the money as a spend policy engine: per-agent and per-period budgets that actually aggregate, counterparty allow lists, human approval thresholds that hold a payment pending instead of declining it, merchant-locked virtual cards for the card rail AgentCore cannot reach, and an immutable audit trail that reconciles to your books rather than to a chain explorer. If your agents pay for APIs over x402, keep AgentCore Payments doing that and put the policy above it. If they also need to buy from vendors, that part was never in scope for AWS.

How AgentCore compares with the other agent control planes

The pattern repeats across every large vendor that shipped an agent control plane in 2026, which is why it is worth naming rather than treating as an AWS quirk. Microsoft’s Agent 365 governs agent identity and access through Entra Agent ID and expresses no monetary amount at all. The accounting and ERP MCP servers govern which tools a model may call against your books, never the amount. The payment MCP servers from Stripe, PayPal, Square and Checkout.com expose write access to live payment APIs and none of them enforces a ceiling. And Salesforce Agentforce models money across 178 standard objects and agents across 41, with no object joining the two, so the largest CRM on the market has nowhere to store an agent budget either.

AWS went further than any of them: it actually shipped a limit. One optional amount, in one currency, for up to eight hours. On the current evidence that makes AgentCore the most complete of the lot, and it still leaves the buying decision ungoverned. The Machine Payment Protocol it now speaks is documented on our MPP page, and the broader question of paying per tool call is covered under MCP payments.

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.

Measured

The complete AgentCore spend-policy surface, measured 9 September 2026

Every property in the published AgentCore API model that can constrain what an agent spends. This is the whole list, not a selection.

Property Type Required What it can express
SessionLimits.maxSpendAmount Amount (value plus currency) Required inside limits A single maximum for one session
Amount.currency Currency enum Required One value only: USD
CreatePaymentSessionRequest.limits SessionLimits Optional Whether a ceiling exists at all
CreatePaymentSessionRequest.expiryTimeInMinutes Integer, 15 to 480 Required How long the session lives, 8 hours maximum
AvailableLimits.availableSpendAmount Amount Read only How much of the session ceiling is left
CryptoX402PaymentInput.permit2AllowanceLimit String, numeric, up to 78 digits Optional An on-chain Permit2 allowance cap
MppPaymentInput.buyerPaysGasFees Boolean Optional Who absorbs the gas fee, not a spend cap
Any per-day, per-month or cumulative total Not present Not applicable Nothing. No such property exists in either model
Any merchant, vendor or counterparty rule Not present Not applicable Nothing. Zero matches across 4,162 properties
Any approval threshold on the payment path Not present Not applicable Nothing. ProcessPayment has no approver or pending state

Rails

What AgentCore Payments can and cannot pay

Taken from the enums in the data-plane model, which are closed sets. If a rail is not an enum member, the API cannot express it.

Rail In the API model Enum evidence What that rules in or out
x402 stablecoin micropayment Yes PaymentType includes CRYPTO_X402 Paid APIs, MCP servers, metered content, other agents
Machine Payment Protocol Yes PaymentType includes MPP Protocol-negotiated machine payments
Embedded crypto wallet Yes PaymentInstrumentType has one member Coinbase CDP or Stripe Privy wallet
USDC balance Yes InstrumentBalanceToken has one member Stablecoin only, on Ethereum or Solana
Corporate or virtual card No No card property exists No merchant checkout, no card statement
ACH, wire or bank transfer No No such property or enum member No supplier payment run
Supplier invoice or purchase order No The string invoice appears zero times No accounts payable workflow
Non-USD currency No Currency enum contains only USD A non-US entity cannot state its ceiling in its own currency

By component

What each AgentCore capability governs, and its published unit price

US list prices from the AWS AgentCore pricing page, verified 9 September 2026. The right-hand column is the point of this table.

Capability Published US price What it governs Can it cap an amount?
Runtime 0.0895 per vCPU-hour, 0.00945 per GB-hour Isolated execution, session state No
Gateway 0.005 per 1,000 API invocations Which APIs become MCP tools, plus rate limits No, only call counts
Identity 0.010 per 1,000 token or API key requests, free via Runtime or Gateway Which credentials the agent may borrow No
Policy 0.000025 per authorization request Whether a given call is authorized No monetary property exists
Memory 0.25 per 1,000 new events, 0.75 per 1,000 records stored monthly What the agent remembers No
Browser 0.0895 per vCPU-hour, 0.00945 per GB-hour Headless browsing No
Code Interpreter 0.0895 per vCPU-hour, 0.00945 per GB-hour Sandboxed code execution No
Agent Registry First 5,000 records free monthly, then 0.400 per 1,000 Which agents are listed and approved for reuse No, approval is registry only
Evaluations 0.0024 per 1,000 input tokens, 1.50 per 1,000 custom evaluations Whether the agent is behaving No
Web Search 7.00 per 1,000 queries Search access No
Payments Pass-through to Coinbase CDP or Stripe Privy, no AWS charge That the agent may pay, up to one session amount One optional USD session maximum

Side by side

AgentCore Payments and a spend policy layer

An honest split. AgentCore owns the rail and the primitive. The policy questions below are the ones its API model has no field for.

Control AgentCore Payments Agentspay
Per-session maximum amount Yes, optional, USD only Yes, and required by policy
Budget that aggregates over a day, week or month No property exists Yes, per agent, team and cost center
Counterparty or merchant allow list No property exists Yes
Human approval above a threshold No, registry approval only Yes, payment held pending a named approver
Card rail for merchant checkout No Yes, merchant-locked virtual cards
Velocity rule on repeated charges No property exists Yes
Non-USD limits No, USD only Yes
Reconciliation to the general ledger Wallet provider ledger, USDC on chain Yes, to your books in dollars
x402 and MPP micropayments Yes, this is what it is for Sits above it as policy, does not replace it
Managed agent runtime Yes, this is what it is for Not our layer, we do not compete here

Frequently asked

Questions people ask about AWS AgentCore

What is AWS AgentCore?

AWS AgentCore, formally Amazon Bedrock AgentCore, is Amazon’s managed platform for running AI agents in production. It bundles thirteen billable capabilities including Runtime for isolated execution, Gateway for turning APIs into MCP tools, Identity for brokering credentials, Policy for authorization, Memory, Browser, Code Interpreter, Observability, Evaluations, Agent Registry and Payments. Pricing is consumption based with no subscription and no minimum.

Does Bedrock AgentCore have spending limits?

Yes, one. AgentCore Payments accepts a per-session maximum spend amount through SessionLimits.maxSpendAmount, expressed as a value and a currency. That single property is the entire spend-policy surface in the published API model. It is optional on CreatePaymentSession, so a session can legally be created without any ceiling, and there is no daily, monthly or cumulative equivalent anywhere in the model.

Can AgentCore Payments use a credit card?

No. The PaymentInstrumentType enum in the AgentCore API model has exactly one member, EMBEDDED_CRYPTO_WALLET, and the balance token enum has one member, USDC. A card can be used to fund the wallet through the provider, but the agent itself pays in stablecoin over x402 or MPP. Nothing in the API can push a charge to a card network, so merchant checkout is out of scope.

Does AgentCore support a daily or monthly budget for an agent?

No. We searched both the AgentCore data-plane and control-plane API models on 9 September 2026 for daily, monthly, weekly, cumulative, aggregate and per-period vocabulary across 4,162 member properties and found no spend-related match. Limits are scoped to a payment session, and a session lasts at most 480 minutes, so each new session begins again at the full maximum.

How long does an AgentCore payment session last?

Between 15 and 480 minutes. The expiryTimeInMinutes field on CreatePaymentSessionRequest is required and constrained to that range, so eight hours is the hard maximum for a single payment session. A long-running agent has to open new sessions, and because no cumulative total exists in the model, each new session gets a fresh spend ceiling.

What currencies does AgentCore Payments support?

One. The Currency enum in the AgentCore API model contains a single member, USD, so a session ceiling can only be stated in US dollars. The value actually moved is USDC, since InstrumentBalanceToken has one member. A company that budgets in euros or pounds cannot express its limit in its own currency through this API.

Is Amazon Bedrock AgentCore Payments generally available?

Yes. AWS announced general availability on 18 August 2026, after a preview that began on 7 May 2026. GA added Machine Payment Protocol support alongside x402, an upto scheme for dynamic pricing, and a Quick Create path for Coinbase wallets. Stripe Privy wallets need credentials from the Privy dashboard.

Which blockchains does AgentCore Payments support?

The wallet network enum lists Ethereum and Solana, and the chain identifier enum lists Base, Base Sepolia, Ethereum, Solana and Solana Devnet. Two of those five are testnets. So in production you are settling USDC on Base, Ethereum or Solana, which is the practical set to plan treasury and reconciliation around.

Does AgentCore Gateway limit how much an agent can spend?

No. Gateway governs which of your APIs and Lambda functions become MCP tools, and it can rate limit them through CreateGatewayRateLimit and BatchPutGatewayRateLimits. Those cap how many times a tool is called, not what a call is worth. Ten calls that each commit 50,000 dollars pass a limit of a hundred calls per minute without objection.

What is the difference between AgentCore Identity and AgentCore Payments?

Identity brokers credentials: it vaults OAuth tokens and API keys so an agent can act against a third-party service without a human pasting a secret. Payments moves money: it gives the agent a stablecoin wallet and settles x402 or MPP charges. Identity decides which systems the agent can reach. Payments decides that it may pay, up to one session amount. Neither decides what it may buy.

Can you require human approval before an AgentCore agent pays?

Not on the payment path. The AgentCore control plane does have ApprovalConfiguration.autoApproval, but it hangs off the Agent Registry and governs whether a published agent is auto-approved for reuse. The ProcessPayment operation takes a payment manager ARN, a session ID, an instrument ID, a payment type and a payment input. It has no approver field, no threshold and no pending state.

How much does AWS AgentCore cost?

It is priced per component with no minimum. Runtime, Browser and Code Interpreter are 0.0895 dollars per vCPU-hour plus 0.00945 dollars per GB-hour of peak memory, and I/O wait is not billed. Gateway is 0.005 dollars per 1,000 invocations, Identity 0.010 dollars per 1,000 credential requests, Policy 0.000025 dollars per authorization, Web Search 7 dollars per 1,000 queries. Payments adds no AWS charge and passes through your wallet provider’s rates.

Do I still need a spend policy layer if I use AgentCore Payments?

If agents only pay for metered APIs and MCP servers with a small session ceiling, AgentCore Payments on its own may be enough. You need a policy layer once any of these is true: the total across a month matters more than the total across eight hours, some counterparties are allowed and others are not, a human must approve above a threshold, purchases happen at merchants that need a card, or finance has to reconcile agent spend to the general ledger in dollars.

Keep reading

More explainers

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

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

AP2 vs ACP vs x402

AP2 vs ACP vs x402

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.

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