Stripe and Tempo just shipped MPP. here's what the Machine Payments Protocol actually standardizes — and what it doesn't.

NOTE: touch is article per recommended_touch. Score 88 ≥85 and product_fit mnemopay qualifies. Flagged needs-human per Dev.to auto-approval cap (0 auto-approved).

Stripe and Tempo just shipped MPP. here's what the Machine Payments Protocol actually standardizes — and what it doesn't.

the Machine Payments Protocol (MPP) is out. co-authored by Stripe and Tempo. the spec: agent initiates payment, network settles, agent receives confirmation. no manual account setup. no credit verification step. no payment method entry.

that's a real advance. it's also exactly one layer of the stack.

here's what MPP standardizes, what it deliberately leaves out, and what you need to build above it before you go to production.

what MPP actually solves

MPP's core contribution is the flow contract: a machine-readable protocol for how an agent announces a payment intent, how the network verifies and routes it, and how the confirmation comes back to the agent in a format it can act on.

before MPP, every agentic payment integration was ad hoc — an agent calling a Stripe function directly, or passing a payment token in a custom header, or triggering a webhook chain designed for humans. MPP standardizes the message format so the agent on one end and the payment network on the other are speaking the same protocol.

that matters a lot for interoperability. it doesn't touch the problems that live inside your agent's authorization scope.

what MPP leaves to you

read the spec carefully and you'll notice what's absent:

spend authorization: MPP defines how the payment flows. it doesn't define who authorized the agent to initiate it, or what the per-invocation cap is, or what happens when the agent exceeds it.
reputation: the protocol has no mechanism for the receiving service to score the paying agent before settling. an agent with a history of failed payments or runaway loops looks identical to a first-time caller.
audit trail: MPP produces a settlement confirmation. it doesn't produce a signed log of which agent version ran the transaction, under what prompt context, with what authorization scope.

those three things are the compliance layer. MPP is the transport layer. they're different.

where MnemoPay fits in an MPP world

MnemoPay (part of BizSuite) is designed to sit above the protocol layer — so as MPP, x402, and AP2 compete and eventually consolidate, the instrumentation above them doesn't change.

what MnemoPay adds to an MPP-enabled stack:

Agent FICO (300-850): before the MPP payment intent fires, the paying agent's score is checked. services set a floor. low-score agents don't initiate.
spend cap enforcement: per-invocation cap is set by the agent's operator before production. if the MPP call would exceed it, the intent is blocked before it hits the network.
signed audit log: every MPP settlement writes a structured log entry with model version, authorization scope, invocation context, and outcome hash. reproducible in 48 hours for compliance review.

672 tests. v1.0.0-beta.1 shipped. 1.4K weekly npm downloads.