Agentspay

Explainer

ServiceNow AI Control Tower and ServiceNow AI Agents: Agentic AI Spend Limits

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.

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

ServiceNow AI Control Tower is ServiceNow's central console for governing AI agents and models across the enterprise, generally available since 6 May 2025 and expanded in May 2026 around five pillars: Discover, Observe, Secure, Govern and Measure. It inventories agents, including third-party ones, runs risk and compliance workflows, and can contain a compromised agent with a kill switch. It does not cap what an agent pays out. In ServiceNow's own SDK, version 4.12.2, the AI agent plugin writes 84 columns and not one of them is an amount, a currency, a budget or a counterparty. Cost appears only as reporting. A per-agent spend limit is a separate control you add underneath it.

What ServiceNow AI Control Tower is, and the question it leaves open

ServiceNow AI Control Tower is the place a ServiceNow customer inventories, governs and measures every AI system the company runs, including agents built outside ServiceNow. ServiceNow announced it and made it generally available at Knowledge 2025 on 6 May 2025, alongside AI Agent Fabric, the communication layer that lets agents talk to other agents and tools over MCP and A2A.

A year later, on 5 May 2026, ServiceNow expanded it around five pillars. Discover finds AI across the estate through 30 new integrations with AWS, Google Cloud and Azure. Observe adds runtime tracing from the Traceloop acquisition. Secure brings identity and access governance from Veza and an AI Gateway for MCP traffic. Govern carries the risk, compliance and lifecycle workflows. Measure, in ServiceNow's words, "provides cost tracking and ROI dashboards." ServiceNow said the expansion would enter its Innovation Lab in May with general availability expected in August 2026.

That is a serious product, and for a company already running ITSM and CMDB on ServiceNow it is the obvious place to put an agent inventory. The question it leaves open is the one finance asks: when an agent under this console tries to pay for something, what refuses the payment? We went to the schema to find out rather than the brochure.

ServiceNow AI agents can already spend money

This is not a hypothetical exposure. ServiceNow ships agents whose documented job is to create orders. The Procurement product recommendation AI agent, per ServiceNow's documentation, "can browse the catalog and create orders" and routes to submission forms based on vendor status and dollar thresholds. The ITAM Purchase order creation agent creates purchase orders. In the developer SDK, one of the tool types an agent can be given maps straight to sc_cat_item, the service catalog item table, which is where ordering happens.

ServiceNow AI agents are built in AI Agent Studio and coordinated by AI Agent Orchestrator, both announced on 29 January 2025 and included at no additional cost for Pro Plus and Enterprise Plus customers at the time. Agents can run in copilot mode, where a person confirms each step, or autopilot, where they do not. The moment one of them runs in autopilot with a catalog or purchasing tool, it is an actor that can commit company money, and the governance question stops being about access and becomes about amount.

The same applies to agents ServiceNow does not own. AI Control Tower discovers agents on Bedrock, AgentCore, Gemini Enterprise, Azure AI Foundry and Salesforce. Many of those can reach a payment MCP server or a corporate card. Seeing them in an inventory is valuable. It is not the same as bounding what they spend.

ServiceNow AI Control Tower pricing: what is public and what is not

ServiceNow does not publish a price list for AI Control Tower, and we found no list price in any ServiceNow press release or product page. Anyone quoting you a precise public number is quoting a partner estimate or a negotiated deal. Here is what can actually be sourced.

On 9 April 2026 ServiceNow announced new packaging that "spans AI assistance, agentic automation, and fully autonomous operations," with AI Control Tower named among the built-in components. The release names no tiers and no prices. Several licensing analysts describe the new tiers as Foundation, Advanced and Prime, replacing the old Standard, Pro, Pro Plus and Enterprise structure, and say AI Control Tower appears in every tier. That comes from secondary sources rather than ServiceNow, so confirm it on your own order form. Fortune reported in May 2026 that ServiceNow was offering AI Control Tower free for a year to qualifying customers, again a secondary report.

The consumption side is better documented. ServiceNow meters AI through assists, and its own documentation describes a 365-day burn-down model. CIO reported in May 2026 that assist usage is difficult for customers to predict. That matters for what follows, because the cost tracking inside AI Control Tower is mostly about this meter, the cost of running AI, not about what an agent spends once it is running.

How we measured what ServiceNow can express about agent spending

We use the same method on every vendor page: read the machine-readable model the vendor publishes, run two vocabulary passes over it, and run the same passes over a part of the same vendor's model that obviously handles money, so we know the method can see money when it is there. ServiceNow does not publish an unauthenticated instance dictionary, so we used the two public surfaces that describe its agent model in detail. Everything below was run on 18 September 2026 without logging in to anything.

Surface one is ServiceNow's own developer SDK, the @servicenow/sdk package family on npm at version 4.12.2, published 15 September 2026. Its AI agent plugin, dist/external-plugins/aiaf.js in @servicenow/sdk-api (179,890 bytes), declares every column an agent definition writes to the instance, across thirteen tables including sn_aia_agent, sn_aia_agent_config, sn_aia_tool, sn_aia_team, sn_aia_usecase, sn_aia_trigger_configuration and sn_aia_version. @servicenow/sdk-core adds 109 typed properties for agent authoring. The same package bundles 185 platform table schemas with 3,477 columns, which gives us a control in the same artifact.

Surface two is the ServiceNow documentation corpus ServiceNow publishes on GitHub for LLM consumption, the ServiceNow/ServiceNowDocs repository on the Australia branch. We scanned 268 AI Control Tower files (117,313 words), all 1,710 intelligent-experiences files (868,629 words), and 1,359 source-to-pay files (543,582 words) as the control.

The money pass looks for amount|currency|budget|spend|price|cost|payment|invoice|monetary|limit|dollar|purchase. The scope pass looks for merchant|vendor|supplier|allowlist|approval|approver|threshold|daily|monthly|cumulative. The claim this supports is precise: there is no spend-limit vocabulary in the shipped, documented agent model. A customer can always add a custom field, and a custom field is exactly the thing you would then have to build, test and enforce yourself.

The finding: 84 AI agent columns, and not one is an amount

The AI agent plugin writes 84 columns to the instance. The money pass matched zero of them. The scope pass matched zero. There is no budget column on sn_aia_agent, no currency on sn_aia_tool, no approval threshold on sn_aia_usecase, no merchant or supplier field anywhere in the agent tables.

The 109 typed properties in the SDK core produced two money hits and one scope hit, and all three are false positives worth naming so nobody misreads them. limit is the maximum number of records a query returns. searchResultsLimit is the same idea for search. documentMatchThreshold is a semantic similarity score between 0 and 1. None of them has a unit of money.

The closest things to a control in the agent model are executionMode, which switches between copilot and autopilot, and max_auto_executions, which counts runs. Both are useful. Neither can express "this agent may spend 2,000 dollars a month with these three suppliers and anything above 500 needs a person." The documentation pass agrees: across 117,313 words of AI Control Tower documentation our cap-phrase search (spend limit, budget cap, per transaction, maximum amount, virtual card and similar) returned zero matches, and so did the full 868,629-word intelligent-experiences corpus.

Where the money words in AI Control Tower actually point

The AI Control Tower documentation does contain money words, 164 of them, a rate of 14.0 per 10,000 words. We read them in context, because a count on its own proves nothing. Every one is about visibility or value, not enforcement.

The clearest example is ServiceNow's own heading, "Control AI spend," which it defines as monitoring AI system costs, including token-level costs across model providers, "so spend is visible and predictable." The rest are project portfolio fields for planned, budgeted and actual cost on AI projects, a Value dashboard that computes amount saved from hours multiplied by an hourly rate, and "max skill calls per execution," which is a usage count. That is a good measurement toolkit for the cost of running AI and the return on it. It is looking at the other side of the ledger from an agent that is about to pay a vendor.

This is the same line we found at Google, where the Gemini Enterprise cap governs the consumption bill, and at Salesforce, where Agentforce Digital Wallet alerts on credits. ServiceNow draws it slightly further toward reporting: it tracks and forecasts the meter, and leaves the payment itself alone.

The two ServiceNow kill switches, and why neither is about money

"ServiceNow AI Control Tower kill switch" is one of the most searched phrases around the product, and it refers to two different features that are easy to confuse.

The AI Control Tower kill switch protocol contains a rogue agent. A threat response policy applies a Deny Resource policy on AWS Bedrock or AgentCore, Gemini Enterprise Agent Platform or ServiceNow itself, disables the agent through the Azure Resource Manager API on Azure AI Foundry, and can optionally have Okta refuse future tokens. It fires on seven threat categories: Agentic Goal Deviation, Data and Model Poisoning, Excessive Agency, Improper Output Handling, Prompt Injection, Sensitive Information Disclosure and System Prompt Leakage. That is a genuinely cross-platform containment capability and ServiceNow deserves credit for it. None of the seven triggers is an amount or a budget, so an agent that spends steadily within its tool permissions never trips it.

The AI Agents trigger kill switch stops runaway triggers that burn assists. Its defaults are 5 fires per record in 24 hours, 25 distinct records and 3 consecutive days, and it ships with kill_switch.mode set to warn_only, so it disables nothing unless an administrator changes it to enforce. It protects your ServiceNow consumption bill. It is not a ceiling on money the agent pays out.

ServiceNow does model a money limit. It is on the approver, not the agent.

The most useful detail in the whole exercise is where a real money ceiling does exist in ServiceNow. In Sourcing and Procurement Operations, the sn_shop_approval_rule table has an Approval limit field of type Currency, documented as the maximum amount that can be approved by the authorized job codes. That is a proper, enforced, dollar-denominated limit.

It is keyed to a job code, meaning a human approver's role, and it runs on a purchase requisition inside ServiceNow procurement. Two consequences follow. First, it does not bound an agent: it bounds the person who approves what the agent requested, which only helps if every agent purchase flows through a requisition. Second, anything the agent pays for outside that flow never meets the rule at all. A SaaS renewal on a corporate card, a metered API bought over x402, an ad top-up, a cloud reservation or a supplier paid through a QuickBooks or NetSuite MCP connector all sit outside sn_shop_approval_rule.

So ServiceNow knows how to model a spending limit, and put it on the procurement object for humans. The agent object, where you would need it for autonomous spend, has none.

The control worked, so the zero is real

A zero from a regex is only meaningful if the same regex fires on money elsewhere in the same source. It does, hard. In the same SDK package, 19 of the 185 bundled tables matched the money pass across 75 columns. ast_contract matched 16 of its 86 columns, including payment_amount, total_cost and monthly_cost. sc_cat_item, the catalog table an agent tool can point at, matched 8 of 77, including price, list_price and recurring_price. alm_asset matched 6 of 74.

In the documentation the contrast is starker. The source-to-pay control corpus hit the money pass at 315.2 per 10,000 words, against 14.0 for AI Control Tower, a gap of more than twenty times, and it returned the cap phrases our agent corpora never did: maximum amount, per transaction, spend cap and virtual card. Its field tables carry 22 fields typed Currency. The AI Control Tower pages carry no structured field tables at all.

ServiceNow models money in depth wherever a human buys something. The agent model is where it is absent, and on the evidence that is a scoping decision rather than an oversight.

What AI Control Tower does well, honestly

This page is not an argument against buying AI Control Tower, and it would be inaccurate to present it as weak. It is, as far as we can tell, the broadest enterprise AI inventory on the market. It discovers agents across Microsoft, Google, AWS, Salesforce and Anthropic connectors and flags shadow AI. It carries risk and compliance content mapped to the EU AI Act and NIST AI RMF. It runs lifecycle approvals so an agent has an owner and a review before it goes live. It adds evaluations and runtime observability, an MCP gateway, and value dashboards, and it plugs into the CMDB and workflow engine most large US enterprises already run on.

If your question is which agents exist, who owns them, what they can reach and whether they have been reviewed, AI Control Tower answers it well. If your question is what a given agent is allowed to spend, with which counterparty, before a person has to sign off, it does not try to answer it. Those are different controls, and a mature program ends up with both.

Where a spend policy layer fits under AI Control Tower

The clean split is this: AI Control Tower governs whether an agent may exist and what it may reach, and a spend policy layer governs what that agent may pay. They sit on top of each other rather than competing, and the policy layer does not need to know which platform built the agent.

In practice that means giving each agent that can move money a per-agent spending limit that aggregates across a day, a week and a month, a merchant-locked virtual card so an agent authorized to renew one vendor cannot pay a different one, an approval threshold that holds a payment for a named person above an amount you choose, and an audit trail that records which agent paid whom, under which rule, in dollars. Those records are the evidence AI Control Tower's risk and compliance workflows can then consume.

The test for whether you need it is short. If your ServiceNow agents triage tickets, summarize incidents and draft knowledge articles, AI Control Tower plus the assists kill switch set to enforce is proportionate, and you can stop here. If any agent, on ServiceNow or anywhere AI Control Tower discovers, has a tool that creates orders, pays invoices or holds a card, you are governing identity and trusting the amount to a prompt. Pick a plan and put a ceiling on the first agent in minutes.

How ServiceNow compares with the other agent control planes

Six vendors, six published models, one pattern. We have measured the machine-readable models behind the major agent control planes shipped in 2026 using the same method each time. Microsoft Agent 365: 48 agent-governance types, 623 properties, zero money. Salesforce Agentforce: 178 money objects and 41 agent objects out of 1,757, with an intersection of zero. AWS AgentCore: one optional per-session USD amount on a session that lives at most eight hours, with no cumulative total. Gemini Enterprise: a real hard cap that refuses work, scoped to a billing project rather than an agent. The payment MCP servers: write access to live payment APIs with no ceiling at all.

ServiceNow is the sixth: 84 agent columns, zero money, a containment kill switch that fires on threats rather than amounts, and the only real money limit in the product attached to a human approver's job code. Every one of these organizations governs identity, access and capability, and every one declines to govern the amount a specific agent may commit to a specific counterparty. For the category as a whole, see AI agent governance and the vendor comparison of AI agent payment platforms.

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.

Timeline

ServiceNow AI Control Tower and AI agents, from ServiceNow's own announcements

Dates from ServiceNow newsroom releases and release notes. ServiceNow publishes no list price for AI Control Tower.

Date What ServiceNow shipped What it governs Money control in it
29 Jan 2025 AI Agent Orchestrator and AI Agent Studio, available March 2025 Building and coordinating agents None
6 May 2025 AI Control Tower, generally available; AI Agent Fabric in early adopter Inventory, governance, agent to agent and tool traffic over MCP and A2A None
9 Apr 2026 New packaging with AI Control Tower built in Licensing across AI assistance and agentic automation Not a control
5 May 2026 Expanded AI Control Tower: Discover, Observe, Secure, Govern, Measure; GA expected August 2026 Cross-cloud discovery, tracing, identity, MCP gateway, cost tracking and ROI Cost tracking and dashboards, no cap
Australia Patch 4 Kill switch protocol for rogue agents Containment on seven threat categories No spend trigger

Measured

ServiceNow's published agent model, measured 18 September 2026

Unauthenticated sources only. The SDK is @servicenow/sdk 4.12.2 on npm; the docs are the ServiceNow/ServiceNowDocs repository, Australia branch.

Source scanned Size Money-vocabulary hits Reading
AI agent plugin columns (sdk-api aiaf.js) 84 columns, 13 sn_aia tables 0 No amount, currency, budget or counterparty on any agent table
AI agent typed properties (sdk-core) 109 properties 2, both record-count limits False positives, no unit of money
AI Control Tower documentation 268 files, 117,313 words 164 (14.0 per 10k words), 0 cap phrases All visibility or ROI, none enforcement
CONTROL: bundled platform tables (same SDK) 185 tables, 3,477 columns 75 columns in 19 tables ast_contract 16 of 86, sc_cat_item 8 of 77
CONTROL: source-to-pay documentation 1,359 files, 543,582 words 17,136 (315.2 per 10k words), 8 cap phrases Money modeled in depth where humans buy

Two kill switches

The ServiceNow kill switches compared

Both are real and both are useful. Neither is triggered by an amount.

Feature What triggers it What it stops Default state
AI Control Tower kill switch protocol Seven threat categories, from Prompt Injection to Excessive Agency The agent, via Deny Resource on Bedrock, AgentCore, Gemini, ServiceNow; ARM on Azure; optional Okta token block Configured per threat response policy
AI Agents trigger kill switch 5 fires per record in 24h, 25 records, 3 consecutive days Runaway triggers consuming assists warn_only, disables nothing until set to enforce
A spend limit on the agent An amount, a counterparty or a cumulative total The single payment that breaks the rule Not present in the shipped agent model

Side by side

What ServiceNow AI Control Tower governs and what a spend policy layer adds

An honest split. ServiceNow owns the inventory, the risk workflows and containment, and we do not compete on any of it.

Control ServiceNow AI Control Tower A spend policy layer
Enterprise inventory of agents, incl. third-party Yes, across AWS, Google Cloud, Azure and more Not its job
Risk, compliance and lifecycle approval of an agent Yes, EU AI Act and NIST AI RMF content Not its job
Containment of a compromised agent Yes, kill switch protocol on seven threat types Revokes the agent's payment credential
Cost of running AI (assists, tokens) Tracked and forecast in Measure Not its job
Hard cap on what an agent pays out No field exists in the agent model Yes, per transaction, day, week and month
Counterparty or merchant rule per agent No; supplier rules exist for human procurement Allow list, deny list, merchant-locked card
Approval above a dollar amount Per approver job code, on requisitions only Per agent, on every payment rail
Spend outside ServiceNow procurement Not visible to the approval rule Yes, card, ACP, AP2 and x402
Audit record of which agent paid whom Logs and traces, not payments Payment-level, agent-attributed, in dollars

Frequently asked

Questions people ask about ServiceNow AI Control Tower

What is ServiceNow AI Control Tower?

ServiceNow AI Control Tower is ServiceNow's central console for governing AI across an enterprise. It inventories AI agents and models, including ones built on AWS, Google Cloud, Azure and Salesforce, runs risk, compliance and lifecycle workflows, observes agents at runtime, and measures cost and return. It became generally available on 6 May 2025 and was expanded in May 2026 around five pillars: Discover, Observe, Secure, Govern and Measure.

How much does ServiceNow AI Control Tower cost?

ServiceNow does not publish a list price. Its April 2026 packaging announcement names AI Control Tower as a built-in component but gives no tiers or prices. Licensing analysts report it is included across the new Foundation, Advanced and Prime tiers, and Fortune reported a free first year for qualifying customers, both secondary sources. AI usage itself is metered separately in assists.

Does ServiceNow AI Control Tower have a kill switch?

Yes. Its kill switch protocol contains a rogue agent by applying a Deny Resource policy on AWS Bedrock, AgentCore, Gemini Enterprise or ServiceNow, disabling it through Azure Resource Manager on Azure AI Foundry, and optionally blocking Okta tokens. It fires on seven threat categories such as Prompt Injection and Excessive Agency. None of those triggers is a spending amount.

What are ServiceNow AI agents?

ServiceNow AI agents are autonomous or supervised agents built in AI Agent Studio and coordinated by AI Agent Orchestrator, which ServiceNow announced on 29 January 2025. They run in copilot mode, where a person confirms each step, or autopilot. ServiceNow ships prebuilt agents for ITSM, HR, customer service and procurement, including agents that create orders and purchase orders.

What is ServiceNow AI Agent Fabric?

AI Agent Fabric is the communication layer ServiceNow announced with AI Control Tower in May 2025. ServiceNow describes it as the backbone for agent to agent, agent to tool and system to system communication, using the MCP and A2A protocols, so ServiceNow agents can work with agents and tools from other vendors. ServiceNow supports A2A version 0.3 for external agents.

Can ServiceNow AI agents create purchase orders?

Yes. ServiceNow documents an ITAM Purchase order creation agent, and a Procurement product recommendation agent that can browse the catalog and create orders, routing by vendor status and dollar thresholds. Those thresholds route the request; they are not a budget on the agent. An agent in autopilot with a catalog tool is an actor that can commit company money.

Can you set a spending limit on a ServiceNow AI agent?

Not in the shipped agent model. In @servicenow/sdk 4.12.2 the AI agent plugin writes 84 columns across 13 sn_aia tables, and none is an amount, currency, budget or counterparty. The nearest controls are copilot versus autopilot and a maximum number of automatic executions. A per-agent spend limit has to be built as a custom field or added as a separate policy layer.

Does AI Control Tower track AI costs?

Yes. The Measure pillar provides cost tracking and ROI dashboards, and ServiceNow describes controlling AI spend as monitoring AI system costs, including token-level costs across model providers, so spend is visible and predictable. That covers the cost of running AI. It does not cap payments an agent makes to vendors, which never reach the AI cost meter.

What are ServiceNow assists?

Assists are ServiceNow's consumption unit for Now Assist and AI agents. Each AI action draws down assists, and ServiceNow documentation describes a 365-day burn-down model. The AI Agents trigger kill switch exists to stop runaway triggers from burning assists, but it ships in warn_only mode and disables nothing until an administrator sets it to enforce.

Does ServiceNow have any money limit at all?

Yes, in procurement. The sn_shop_approval_rule table has an Approval limit field of type Currency, the maximum amount the authorized job codes can approve. It is keyed to a human approver's role and runs on purchase requisitions, so it bounds the approver rather than the agent, and never sees spend that happens outside ServiceNow procurement.

Does ServiceNow support MCP and A2A?

Yes. ServiceNow supports A2A version 0.3 for external agents, ships an MCP client and server with an MCP catalog, and added an AI Gateway that proxies MCP traffic, listed in AI Control Tower from the September 2026 release. ServiceNow's SDK documentation says MCP connections are approved in AI Control Tower, which governs the connection rather than the payments made through it.

What are the ServiceNow AI Control Tower competitors?

For agent governance the closest alternatives are the platform consoles from the big clouds and suites: Microsoft Agent 365, Salesforce Agentforce governance, AWS AgentCore and Gemini Enterprise. Each governs identity, access and capability in its own ecosystem, while AI Control Tower spans several. None of them, in our schema measurements, caps what an individual agent may spend.

Do I need a spend policy layer if I use ServiceNow AI Control Tower?

Only if your agents can move money. If they triage tickets, summarize incidents and draft articles, AI Control Tower with the assists kill switch set to enforce is proportionate. If any agent can create orders, pay invoices, hold a card or call a paid API, you need a per-agent budget, a counterparty rule and an approval threshold enforced before the payment.

Keep reading

More explainers

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

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