Machine Payments Protocol Explained [What It Changes]

Machine Payments Protocol (MPP) is an open, internet-native standard built on HTTP 402 that enables AI agents to autonomously pay for services, APIs, and digital resources in real time — without human approval at each transaction.

TL;DR

• Stripe and Tempo launched the Machine Payments Protocol on March 18, 2026

• Built on the long-reserved HTTP 402 status code ("Payment Required"), now repurposed for machine-initiated transactions

• Over 100 services were listed in the MPP directory at launch, spanning model providers, developer infrastructure, compute, and data services

• Agents transact via stablecoins, credit/debit cards, or buy-now-pay-later through Shared Payment Tokens

• The protocol is open: any service can implement it, not locked to Stripe infrastructure

• SMBs gain AI-driven procurement and pay-per-use API orchestration without manual checkout flows

What Happened and Why Now

According to Fortune, Stripe and Tempo launched the Machine Payments Protocol on March 18, 2026 — an open-source network that supports payments in both fiat and cryptocurrency and is compatible with Stripe's existing AI payments infrastructure. The "Payment Required" status code itself — HTTP 402, reserved in the HTTP specification for decades but never formally implemented for the web — is most directly invoked by Coinbase's separate x402 network, which Fortune describes as "a callback to a message encoded by early internet pioneers that returned a 402, or 'Payment Required,' error." According to Stripe's blog, MPP lets AI agents pay for services, APIs, and digital resources in real time without human intervention.

According to Forrester, the protocol signals a turning point for micropayments. The economics are why. Standard card pricing carries a roughly $0.30 fixed fee plus 2.9% per transaction (see Stripe's pricing), so sub-cent API calls are structurally unviable on card rails — whereas MPP's stablecoin rail carries no fixed per-transaction floor. The ability to pay fractional amounts per API call or per document processed at machine speed and scale is what makes autonomous agent workflows economically practical.

The constraint that broke is the human checkout loop. When an AI agent encounters a paid API or a service wall mid-task, the only resolution path has been a human approval step — add payment method, confirm billing, re-authorize. That friction makes fully autonomous multi-step agent workflows impractical for any task that requires accessing paid external resources.

According to Techstrong, over 100 services were listed in the MPP directory at launch, spanning model providers, developer infrastructure, compute platforms, and data services. That immediate network effect at launch — rather than gradual rollout — reflects a coordinated standard adoption rather than a single-vendor feature.

How the Machine Payments Protocol Works

The HTTP 402 Foundation

HTTP 402 was reserved in the original HTTP/1.1 specification as "Payment Required" but never formally standardized. Browser interactions never triggered it. MPP takes that reserved code and gives it a machine-readable specification: when an agent hits a paid resource, the server returns 402 with a structured payment object. The agent reads the object, selects a payment method from its Shared Payment Token, executes the transaction, and retries the request — all without surfacing a human payment UI.

The mechanism in plain language: the agent encounters a "this costs something" signal, pays using its pre-authorized credentials, and continues the task. No checkout page, no human confirmation, no session break.

Shared Payment Tokens

According to Stripe's blog, agents transact through Shared Payment Tokens — pre-issued credentials covering 3 payment methods (stablecoins, cards, BNPL) that encapsulate the agent's authorized payment scope. The token travels with the agent's session, not with a specific service, so a single token can authorize payments across participating services without per-service payment setup.

According to Stripe, MPP supports 3 payment rails — stablecoins, credit/debit cards, and BNPL — through Shared Payment Tokens. The multi-method flexibility matters for international SMBs and scenarios where stablecoin settlement reduces FX friction.

The Open Standard Architecture

MPP is not proprietary to Stripe. Any service can implement the HTTP 402 response object following the open specification. Stripe's participation means existing Stripe-integrated businesses have a natural on-ramp, but the protocol's openness means a competitor payment infrastructure could implement the same standard.

Timeline: From HTTP 402 to Live Protocol

Sources: Stripe; Forrester.

MPP vs. Traditional Card Processing: Cost Structure

Traditional card infrastructure sets a hard floor on micropayments. A standard Stripe card processing fee runs approximately 2.9% + $0.30 per transaction — meaning a $0.10 API call costs more in processing fees than the call itself, making sub-dollar pay-per-use economically unviable. According to WorkOS, the Tempo chain underlying MPP denominates fees in stablecoins, designed to be negligible, with no fixed per-transaction floor — making $0.01–$0.10 API calls economically viable at agent scale in a way that card rails cannot.

Standard Stripe card processing runs ~2.9% + $0.30 per transaction, making sub-dollar API calls cost more in fees than the call itself — the exact problem MPP's stablecoin rail solves, per WorkOS.

Sources: WorkOS; Stripe.

The AP automation market reached USD 6.94 billion in 2026 — the infrastructure investment wave that MPP's machine-native payment rail now connects to, per Dokka.

What MPP Actually Changes for SMBs

Before MPP, AI agent workflows that needed to pay for anything — a data API, a compute resource, a premium SaaS feature — hit a hard wall. The agent could not proceed without human authorization of a payment. That made "fully autonomous" an exaggeration for any real-world business workflow touching paid external services.

MPP removes that wall. The practical implications by use case:

Pay-per-use API access: An agent running a market research workflow can query a paid data API mid-task, pay per call via its Shared Payment Token, and return the results — no pre-purchase of an API subscription required.

AI-driven procurement: An agent authorized to source and purchase digital resources can complete the purchase step without escalating to a human approver for each transaction, provided the spend stays within the token's pre-authorized limits.

Automated operational spend: Subscription renewals, overage charges, and per-seat licensing events can be handled by an agent as they trigger, rather than queuing for a human billing review cycle.

According to Forrester, the protocol represents a meaningful infrastructure shift for micropayment viability — the economics of sub-dollar per-call transactions become viable when the authorization overhead is removed. Forrester frames this alongside the broader AP automation wave: 66% of AP teams still manually key invoices into ERP systems, a bottleneck that machine-initiated payments are now positioned to close end-to-end.

Capability Comparison: Manual Checkout vs MPP Agent Flow

Sources: Stripe; Forrester.

Who Adopted MPP at Launch

According to Techstrong, the 100+ launch-day directory spans model providers, developer infrastructure, compute platforms, and data services, with named early services including Browserbase, PostalForm, and Stripe Climate. The infrastructure coverage matters: if the AI models agents run on and the commerce layer they transact in both implement MPP, the protocol has a realistic path to ubiquity in agent-driven workflows.

The AP automation market is valued at USD 6.94 billion in 2026, projected to reach USD 12.46 billion by 2031 — per Dokka citing Mordor Intelligence — the market MPP's payment rail now targets directly.

Source: Techstrong.

Where Automation Orchestration Fits

MPP solves the payment step in an agent workflow. It does not solve orchestration — routing which agent does what, managing multi-agent handoffs, handling errors and retries across the full task chain.

Teams already running orchestrated workflows through US Tech Automations can treat MPP as a payment capability layer that slots into existing agent architecture. An agent in a procurement workflow gains the ability to execute payment steps inline; the orchestration logic — what triggers the procurement task, how the result routes to approval and accounting — stays in the orchestration layer.

For accounting-specific implications of MPP — including accounts payable automation and client billing workflows — see What Machine Payments Protocol Means for Accounting Firms.

For broader SMB operational workflows incorporating autonomous agent payments, see What Machine Payments Protocol Means for Small Businesses.

The orchestration layer that connects MPP-capable agents to existing ERP, accounting, and operations systems is where US Tech Automations operates — the agentic workflow platform handles the multi-step coordination that MPP's payment step plugs into.

Signal vs Speculation

Demonstrated facts (sourced, as of June 2026 — Stripe):

• MPP launched March 18, 2026, announced by Stripe and Tempo as an open standard — confirmed by Fortune

• Over 100 services were listed in the MPP directory at launch, spanning model providers, developer infrastructure, compute, and data services — confirmed by Techstrong

• Agents transact via stablecoins, cards, or BNPL through Shared Payment Tokens — confirmed by Stripe

• Forrester characterized MPP as a turning point for micropayments — per Forrester

Our read (forecast — not fact):

HTTP 402's adoption as a machine payment signal is a one-way door. Once major AI infrastructure providers and major commerce platforms implement it, any AI agent built on those providers has a payment rail available. The 100+ launch-day directory entries give MPP a network density that is very difficult for a competing standard to overcome in the 12-24 month window.

The risk for SMBs is not adoption — it is governance. Shared Payment Tokens with pre-authorized spending limits introduce a new category of enterprise risk: a misconfigured agent with a token could execute unauthorized purchases at scale before a human notices. The spend control and audit trail architecture around MPP implementation matters as much as the protocol itself.

In 24-36 months, expect MPP-aware orchestration to be a baseline requirement for enterprise AI agent deployments — the same way OAuth became a baseline for API authentication. The firms that build their agent architectures with MPP payment rails now, alongside the governance controls, will have a structural advantage when autonomous procurement becomes routine.

For marketing agencies implications, see What Machine Payments Protocol Means for Marketing Agencies.

Key Takeaways

• Machine Payments Protocol launched March 18, 2026, built on HTTP 402, enabling AI agents to pay for services autonomously in real time

• Over 100 services were listed in the MPP directory at launch, spanning model providers, developer infrastructure, compute, and data services — the network density is already significant

• Agents use Shared Payment Tokens covering stablecoins, cards, and BNPL, pre-authorized within spending limits set by the business

• The protocol is open — any service can implement it, not Stripe-exclusive

• MPP solves the payment step in agent workflows; orchestration across multi-agent tasks requires a separate layer

• The governance risk — misconfigured agents spending against a pre-authorized token — is the primary implementation challenge for SMBs

• As of June 2026, MPP is live and deployed; SMB implementation timelines depend on tool and platform adoption

Frequently Asked Questions

What is Machine Payments Protocol?

Machine Payments Protocol (MPP) is an open internet standard built on HTTP 402 that allows AI agents to pay for services, APIs, and digital resources autonomously — without human intervention at each transaction. It was launched by Stripe and Tempo on March 18, 2026.

What is HTTP 402 and why does it matter?

HTTP 402 is a status code reserved in the HTTP specification since the HTTP/1.1 era as "Payment Required" but never formally implemented for web use. MPP gives it a machine-readable specification, so servers can signal "this resource requires payment" to AI agents in a standardized way.

What payment methods does MPP support?

According to Stripe, MPP supports stablecoins, credit/debit cards, and buy-now-pay-later (BNPL) transactions through Shared Payment Tokens.

Is MPP exclusive to Stripe?

No. MPP is an open protocol — any service can implement the HTTP 402 response specification. Stripe's involvement means existing Stripe-integrated businesses have a natural adoption path, but the standard is not proprietary.

How do businesses control agent spending under MPP?

Shared Payment Tokens carry pre-authorized spending limits set by the business. Agents can only transact within those limits. Governance architecture — how limits are set, monitored, and audited — is a business implementation responsibility, not a protocol feature.

What does MPP mean for accounts payable automation?

MPP enables agents to handle payment steps in AP workflows without human checkout authorization. For accounting-specific workflow implications, see What Machine Payments Protocol Means for Accounting Firms.

Information current as of June 2026. Protocol capabilities, adopter list, and SMB implementation guidance may evolve; verify current state at stripe.com/blog/machine-payments-protocol.

Explore how to build MPP-capable agent workflows with the agentic workflow platform.