What Is MPP?
MPP (Managed Payments Profile) is Stripe's payment instrument designed for agent-mediated commerce. It is not a new payment rail. It is a managed configuration layer on top of existing card-and-bank infrastructure, designed to let AI agents transact within user-defined constraints (spending caps, category restrictions, time bounds) without the user having to authorise every transaction interactively.
Stripe shipped MPP alongside its broader agent commerce push in 2025, paired with the Tempo blockchain announcement and the agent-toolkit SDK. As of April 2026, MPP is the primary agent-payment offering from Stripe and the most operationally mature card-rail-based agent payment instrument in production.
Why MPP Matters
The friction in agent commerce on traditional card rails is interactive authentication. Every transaction triggering 3-D Secure, a one-time-password, or an explicit user confirmation breaks the agentic flow. Cards work for human checkout because humans tolerate the auth step. They do not work cleanly for agents because agents cannot reliably complete interactive auth.
MPP solves this by moving the authorisation upstream: the user creates a managed profile with explicit constraints, and any transaction within those constraints proceeds without further user interaction. The constraints (spending cap per period, allowed merchant categories, blocked merchants, time bounds) become the auth signal. This is the card-rail equivalent of giving an agent a corporate card with a spending limit, and it is what makes card-rail agent commerce operationally viable at scale.
How MPP Works
The user creates an MPP via Stripe's API or hosted UI, attaching it to a payment method (card, bank account, or wallet). The profile expresses constraints: caps, categories, time bounds, merchant allow- and deny-lists. The agent receives credentials tied to the MPP, not the underlying card. When the agent transacts, the transaction is evaluated against the MPP constraints; if it satisfies them, the transaction proceeds; if not, it is rejected and surfaced to the user for explicit authorisation.
Critically, the MPP credentials are revocable, narrowly scoped, and observable. The user can see every transaction the agent attempted, which constraints applied, and which transactions were declined. This audit trail is what regulated industries (fintech, healthcare, legal) need to allow agent commerce at all.
In Practice
For consumers, MPP is invisible: it operates behind the scenes when an agent like ChatGPT Operator or a Stripe-integrated commerce agent makes a purchase on the user's behalf. For merchants, MPP transactions look like standard card transactions with additional metadata indicating the agent context. For agent platforms, MPP is the Stripe-recommended way to handle agent-mediated card payments without the interactive-auth friction.
For publishers and brands, the implication is that card-rail agent commerce works through MPP-style instruments, not through standard cards. Pricing, refund policies, and dispute handling all need to be designed for the agent-mediated case, where the buyer of record is the user but the transactor is an agent operating under managed constraints. This is a meaningful operational change from human-checkout commerce, and merchants who instrument for it cleanly see better conversion in agent-mediated flows.
Commonly Confused With
MPP is not a competitor to AP2 or x402. MPP is a Stripe-specific implementation that operates on card and bank rails. AP2 is a cross-rail protocol that could sit above MPP. x402 is HTTP-level payment-required signalling. The three layers compose: an x402 response could trigger an AP2 authorisation flow that ultimately settles through MPP on Stripe.
