AP2, x402, MPP — three payment protocols in one year means one thing: the normalization layer is the product
PYMNTS documented what's happening in agentic payments in 2026 with unusual clarity: three competing open payment protocols launched in the same year. AP2 led by Google. x402 developed by Coinbase. MPP co-authored by Stripe and Tempo.
when Stripe ships an open protocol for a market, that market is real. Stripe doesn't experiment publicly on speculative markets — they ship when they've seen enough customer signal to validate the investment. MPP is Stripe saying: agent payments are a product category we're taking seriously.
the problem isn't that Stripe is wrong. the problem is that Google and Coinbase also shipped, and none of them has cleared the field. that's the condition builders actually have to work in.
three protocols is not the same as one standard
an open protocol becomes a standard when adoption reaches the point where it's more expensive to build against something else than to conform. IETF standards, HTTP, TLS — these are standards because the ecosystem coalesced around them. AP2, x402, and MPP are each trying to become that standard, but none of them is there yet.
for a developer shipping a payment-gated MCP server or agentic application today, this creates a real decision tree. the server needs to pick a settlement rail. every rail it picks excludes callers on the other two. and none of the three has enough market share to justify betting that it wins.
the rational behavior in a fragmented protocol market is to not bet on a protocol — to build against an abstraction that handles whatever protocol the caller comes in on, and normalize it to the settlement rail the server prefers.
what the normalization layer actually needs to do
the three protocols have different authorization models, different settlement mechanics, and different cryptographic primitives. AP2 uses signed mandates. x402 extends the HTTP 402 payment-required response. MPP (Stripe/Tempo) is built on Stripe's existing auth infrastructure.
normalizing across all three means:
- inbound protocol detection — identify which protocol the calling agent is using without requiring the caller to declare it explicitly
- auth translation — convert the calling protocol's authorization primitive (signed mandate, 402 token, Stripe payment intent) into the server's preferred auth representation
- settlement routing — route the actual settlement to the server's preferred rail, independent of what the caller used for auth
- per-call receipt generation — produce a tamper-evident receipt per invocation that records which protocol was used, what was authorized, and what settled, for both parties' audit trails
MnemoPay handles all four of those layers. 672 tests, v1.0.0-beta.1, 1.4K weekly npm downloads. the normalization problem is the one we built for specifically, because we saw the fragmentation coming when AP2 and x402 both shipped before MPP and the ecosystem clearly wasn't consolidating.
why Stripe's entry changes the picture
Stripe co-authoring MPP is the most significant of the three launches, not because MPP is technically superior (the jury is out), but because Stripe's distribution changes the adoption curve. every developer who already uses Stripe for their current product has a direct path to MPP adoption with minimal integration work. that's a moat that AP2 and x402 have to overcome.
but Stripe's distribution advantage for server-side developers doesn't solve the problem for callers. a Claude or GPT-4o agent running in a multi-platform context might be authorized under AP2 by a Google Cloud principal, or under x402 by a Coinbase-integrated wallet, regardless of what the server prefers. the caller's protocol context isn't chosen by the server developer.
the market is going to be multi-protocol for the foreseeable future. the normalization layer isn't a transitional product that becomes unnecessary when consolidation happens — it's the infrastructure that makes consolidation possible without forcing every server developer to rebuild their payment integration for the winning protocol.
Top comments (0)