Agentspay

Explainer

AI Agent Cost: AI Agent Pricing Models, Cost Per Month, and the Line Nobody Budgets

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.

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 AI agent costs money in three places, and most budgets only cover two. You pay to build or license it, you pay to run it (model tokens, infrastructure, and per-action or per-outcome vendor fees), and then the agent spends money on your behalf: the purchases it makes, the APIs it calls, the subscriptions it renews. Published vendor rates pin the second line down precisely, with Salesforce billing Agentforce actions at $0.10 each and Intercom billing Fin from $0.99 per resolved outcome. No vendor prices the third line, because it is not their money. The first two lines are capped by a contract. The third is capped by nothing unless you cap it yourself.

What does an AI agent cost?

There is no single number, and any guide that gives you one is selling something. What there is, reliably, is a structure: build or license, run, and spend. The first two are quotable because a vendor or a contractor owns them. Licensed platforms bill on the models in the table below, custom agents bill on tokens and infrastructure, and a build project bills on engineering time. Those numbers vary by an order of magnitude across the market, which is why published ranges are close to useless for planning: a support deflection bot and a multi-agent procurement system are not the same purchase. What travels across every deployment is the shape of the bill, and the fact that the shape has three parts while most budget spreadsheets have two.

The three lines in an AI agent budget

Build or license is what you pay to have the agent exist: a platform subscription, a development project, or both when you license a framework and build on top of it. It is usually a one-time or annual number and it is the one everybody remembers. Run is what you pay for the agent to do work: model tokens, hosting, vector storage, observability, and the vendor's own metering, whether that is per action, per conversation or per outcome. It is variable, it scales with usage, and it is where most first-year overruns show up. Spend is what the agent pays out to third parties while doing its job. A procurement agent buys supplies. A travel agent books flights. An ad agent places media. A research agent calls metered APIs. That money leaves your account through a payment credential you issued, not through a vendor invoice, so it usually lands in a completely different part of the general ledger from the first two lines. Teams routinely discover it at month end.

AI agent pricing models: how vendors actually bill

Six models cover almost the entire market in 2026, and the industry has been drifting steadily from seats toward consumption. The useful research here is the Poyar State of B2B SaaS and AI Monetization survey of more than 230 companies, run in April and May 2026, which found hybrid pricing rising from 25% to 37% over twelve months. Worth flagging because the number gets misquoted constantly: a widely repeated "43% to 61%" figure does not appear in that research. The practical consequence of the drift is that your AI agent cost is increasingly a function of how much the agent works rather than how many people are on the account, which is better value when volume is low and considerably worse when an agent gets stuck in a retry loop. Read the models below for what each one does to your budget when things go wrong, not just when they go right.

What published AI agent pricing actually says

Most vendors in this category do not publish a rate card, so the honest table is short. Salesforce and Intercom do publish, and their numbers are worth knowing because they anchor the two dominant consumption models: per action and per outcome. Everything else in the market tends to be quoted per deployment. When a comparison article gives you precise pricing for a vendor that publishes none, that pricing was estimated, and you should treat it as such. We have deliberately left estimates out of the table below and marked the gaps as gaps. Verify anything here against your own order form before you build a budget on it, because rate cards in this category are moving quickly.

AI agent cost per month: building a number you can defend

Work the three lines separately and resist the urge to produce one figure. For license, take the annual contract and divide. For run, estimate the unit of work the vendor meters, multiply by expected monthly volume, then multiply again by a retry factor, because agents fail and re-attempt in ways deterministic software does not. If your vendor meters actions, remember a single user request can be many actions: a plan, several tool calls, a summary. For spend, do not estimate at all. Set it. The agent's outbound spend is the only line of the three you can fix in advance by policy, so make it a decision rather than a forecast: this agent may spend up to this amount, at these merchants, this month. That converts your most volatile line into your most predictable one, which is the opposite of how most teams treat it.

Why are AI agents more expensive than the sticker price?

Three reasons, in rising order of how much damage they do. First, the unit is smaller than you think: per-action and per-token billing meter the agent's internal reasoning, not the user's request, so a single question can bill as a dozen events. Second, failure is billable. A retry, a bad plan and a tool call that returns nothing all cost the same as work that succeeded, and agents retry far more than traditional software. Third, and largest, the outbound spend is not in the comparison at all. A team that agonizes over a $0.10 per action rate will hand the same agent a payment credential with no ceiling on it, and the second decision is worth orders of magnitude more than the first. If an agent runs ten thousand actions in a month, that is a thousand dollars of platform cost. If the same agent makes one bad purchasing decision at scale, it can be more than the entire platform contract.

The third line: what your agent spends on your behalf

This is the part of AI agent cost that no pricing page covers, and it is structurally different from the other two. Your platform bill has a natural ceiling, because you signed a contract and the vendor cannot bill you for work you did not ask for. Your token bill has a soft ceiling, because rate limits and context windows bound it. Your agent's purchasing has no ceiling by default. It is bounded only by the credential you gave it and whatever the merchant will authorize. This matters more every quarter because the industry is actively removing the last human checkpoint: agentic checkout is designed to complete a purchase without anyone reading the total, and neither the ACP nor the UCP protocol carries a buyer budget. Scoped payment tokens cap one transaction, not a cumulative total, so an agent can place forty individually sensible orders that add up to a number nobody would have approved. The checkout screen used to be an informal budget control. It is being deleted on purpose.

How to cap the third line before it becomes the whole budget

Outbound spend has to be controlled at the payment instrument rather than in the prompt, because a prompt is a request and an authorization decline is an answer. In practice that means four controls. A funded wallet with a hard cap, so your worst case is a number you chose in advance rather than one you discover. A merchant and category allowlist, so an agent that finds a cheaper unapproved supplier simply fails the authorization instead of succeeding at the wrong thing. An approval threshold, above which the payment pauses and a named person gets a one-tap approve or deny, which puts a human back exactly where the money justifies the latency and nowhere else. And an audit trail that ties every authorization to the agent, its human owner, the task and the policy verdict, because reconciling agent spend after the fact is genuinely hard when there was no browser session and no expense report. That is what Agentspay does across card rails and stablecoins from one control plane. You can see the surfaces on agent spend controls, human approvals and the agent audit trail, and the console at the top of this page runs the same policy engine.

Are AI agents worth it?

Often, but the comparison most teams run is the wrong one. Measuring platform cost against a salary flatters the agent, because the salary is fully loaded and the agent cost is not. A fair comparison puts all three lines against the human total, including the purchases a person would have made anyway, and then adds the cost of the controls and review the agent needs and the person did not. Agents win decisively on volume, on tasks that are tedious and countable, and on work that has to happen at three in the morning. They win less clearly on judgment-heavy work with expensive failure modes, which is precisely the category where uncapped outbound spend does damage. We work through that comparison properly in AI agent cost vs salary. If you are on the other side of this question and trying to price an agent you built rather than budget for one you bought, AI agent monetization covers that, and what agent payment infrastructure costs covers the rail fees specifically.

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.

Pricing models

The six ways AI agents are billed, and what each does to your budget

Hybrid pricing rose from 25% to 37% of B2B AI companies over twelve months (Poyar, State of B2B SaaS and AI Monetization, April to May 2026, 230+ companies). The last row is on no vendor invoice.

Model How you are billed Best when Where the budget breaks
Per seat Flat fee per human user, per month The agent assists a fixed, known team You pay for seats whether the agent works or not
Per action or step Each discrete action the agent takes Workflow agents with countable steps One user request can be a dozen billed actions
Per conversation Each session, regardless of length Support deflection at steady volume A traffic spike is a cost spike, one to one
Per outcome or resolution Only when the agent resolves the task You want to pay for results, not effort Everything depends on how "outcome" is defined
Per token Model input and output tokens Custom agents on raw model APIs Long context and tool chatter dominate the bill
Hybrid: platform plus usage Base fee plus consumption on top Most enterprise contracts in 2026 The base is quoted, the variable is where overruns live
The agent's own spend Whatever the agent buys from third parties Any agent holding a payment credential Uncapped by default, and on no vendor invoice

Published rates

What AI agent vendors actually publish, verified August 2026

Only figures published by the vendor itself are listed. Gaps are marked as gaps rather than filled with third-party estimates. Confirm against your own order form before budgeting.

Vendor What it meters Published rate Notes
Salesforce Agentforce Actions, via Flex Credits $0.10 per standard action (20 credits), $0.15 per voice action (30 credits) Credits sold at $500 per 100,000. A $2 per conversation model is also offered
Intercom Fin Resolved outcomes From $0.99 per outcome Plus seats: Essential $29, Advanced $85, Expert $132 per seat per month, billed annually
Zendesk Automated resolutions Not published Zendesk meters automated resolutions but publishes no per-resolution rate
Sierra Outcomes Not published Third-party per-outcome estimates circulate widely and are not confirmed by Sierra
OpenAI and Anthropic models Input and output tokens Per million tokens, published per model The floor under any custom-built agent
Your agent's purchases Nothing meters this Not applicable No vendor invoices it, no contract caps it. This is the line this page exists to flag

Frequently asked

Questions people ask about AI Agent Cost

How much does an AI agent cost?

There is no single figure, because a support deflection bot and a multi-agent procurement system are different purchases. What is consistent is the structure: you pay to build or license the agent, you pay to run it on tokens and per-action or per-outcome fees, and the agent then spends money on your behalf.

How much does an AI agent cost per month?

Build the monthly number from three separate lines rather than one. Divide the annual license by twelve, estimate run cost as metered units times monthly volume times a retry factor, and then set outbound spend by policy instead of forecasting it. The third line is the only one you can fix in advance.

How much does an AI agent cost to run?

Running cost is model tokens plus infrastructure plus whatever the vendor meters. Salesforce publishes $0.10 per standard Agentforce action and Intercom publishes from $0.99 per resolved Fin outcome. Budget above your naive estimate, because retries, failed tool calls and internal reasoning steps are all billable events.

Do AI agents cost money?

Yes, in three ways. There is a build or license cost to have the agent at all, a running cost that scales with how much work it does, and the money the agent itself pays out to third parties. Free tiers generally cover the first two at low volume and never touch the third.

Why are AI agents so expensive?

Mostly because the billed unit is smaller than people expect and failure is billable. Per-action and per-token pricing meters the agent's internal reasoning, not the user's request, so one question becomes many events. Retries cost the same as successes. The largest cost, outbound spend, is usually left out of the comparison entirely.

Are AI agents expensive to run?

Relative to traditional software, yes, because consumption pricing removes the flat-fee ceiling that made SaaS budgets predictable. Relative to the labor they replace, usually no. The risk is not the unit rate but the variance: an agent stuck in a retry loop or buying from an unapproved supplier costs far more than its rate card suggests.

Are AI agents worth it?

Usually on high-volume, countable, tedious work, and less clearly on judgment-heavy tasks with expensive failure modes. The comparison only works if you count all three cost lines against the fully loaded human alternative, including the controls and review an agent needs that a trained employee does not.

What are the AI agent pricing models?

Six dominate: per seat, per action or step, per conversation, per outcome or resolution, per token, and hybrid base-plus-usage. Hybrid rose from 25% to 37% of B2B AI companies in twelve months. A seventh line, the agent's own outbound spend, appears on no vendor rate card at all.

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

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

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