April 7, 2026
The Protocol War That Doesn't Matter
On March 18, Stripe launched a blockchain. Tempo went live on mainnet with roughly 10,000 transactions per second, sub-second finality, and a partner list that includes Visa, Mastercard, Deutsche Bank, OpenAI, Anthropic, and DoorDash. Alongside it shipped the Machine Payments Protocol, an open standard for agent-to-machine payments co-authored by Stripe and Tempo Labs and submitted to the IETF as a formal internet draft.
MPP uses HTTP 402. So does x402, the protocol Coinbase created and transferred to the Linux Foundation last year with backing from Google, Stripe, and Visa. Both protocols solve the same problem: when an AI agent hits an API, the server responds with a 402 Payment Required status code, the agent pays, and the server delivers the data. No API keys. No accounts. No invoices. The payment is the authentication.
The crypto press and the VC world have framed this as a binary contest. x402 or MPP. Coinbase or Stripe. Pick a side.
That framing is wrong, and building a business around it is a mistake.
What actually differs between them
The core payment flow is nearly identical. Agent requests resource. Server returns 402 with payment terms. Agent signs a payment. Server verifies, settles, delivers. Both protocols accomplish this over plain HTTP.
MPP adds three things that x402 does not have natively.
Sessions. MPP introduces payment channels where an agent deposits funds into an on-chain escrow, then streams signed vouchers off-chain as it consumes a service. The server verifies each voucher with a single signature check that takes microseconds. When the session ends, one on-chain transaction settles the final amount and returns unused funds. Two blockchain transactions total, regardless of whether ten or ten thousand individual payments occurred in between. Tempo's documentation describes this as "OAuth for money." For use cases like LLM token streaming, real-time data feeds, and multi-step agent workflows, sessions are a meaningful architectural improvement over per-request settlement.
Multi-rail payments. MPP natively supports stablecoins on Tempo, credit cards and wallets through Stripe, and Bitcoin over the Lightning Network through Lightspark. Visa has extended it for card payments on their network. x402 is stablecoin-only today, though the protocol itself does not preclude other payment methods.
x402 compatibility. This detail gets buried in the coverage. MPP can express x402 payments. An x402 server can accept MPP-formatted payments without code changes. The reverse is not true; x402 does not understand MPP sessions or fiat rails. MPP was designed as a superset, not a replacement.
x402 retains its own advantages. It is genuinely permissionless and chain-agnostic, running today on Solana, Base, Polygon, Arbitrum, Optimism, Avalanche, SKALE, and BSC through various facilitators. MPP runs primarily on Tempo's single L1. x402 requires no ecosystem buy-in. MPP's smoothest path runs through Stripe's platform and Tempo's chain.
These are real differences. They are also not the differences that will determine which protocol dominates.
Cloudflare is the kingmaker
David Christopher's Bankless analysis published on March 21 makes the case clearly. Cloudflare sits between websites and the traffic that hits them. About 20 percent of the internet runs through Cloudflare's infrastructure. They already operate pay-per-crawl, a system that charges AI bots micropayments for access to websites rather than blocking them outright. When a bot hits a Cloudflare-protected page, it pays or it gets a 402.
Cloudflare co-founded the x402 Foundation with Coinbase in September 2025. Days later they announced the NET Dollar, a stablecoin purpose-built for agentic payments. They have not named a stablecoin issuer. In March 2026, a report from The Information identified Coinbase and ZeroHash as companies competing for that deal. The phrasing left room for Stripe.
The day after Tempo's mainnet launch, Cloudflare released an MPP compatibility proxy. They are hedging.
Whoever issues the NET Dollar gets their payment standard prioritized across a fifth of the internet. If Coinbase issues it, Cloudflare leans into x402. If Stripe issues it, MPP gets the tailwind. The protocol debate is downstream of a business development negotiation happening in private between three companies.
Building a payments business that depends on the outcome of that negotiation is building on sand.
The settlement layer is the business
Protocols are specifications. Facilitators are businesses. This distinction matters enormously and gets lost in protocol discourse.
When an agent makes an x402 payment, the protocol defines the HTTP headers and the payment format. But the actual work of verifying that payment, co-signing the transaction, sponsoring gas fees, broadcasting to a blockchain, confirming settlement, and generating a receipt is done by a facilitator. The facilitator is a live service with fee payer wallets, RPC connections, security validation, rate limiting, and operational overhead. Protocols are free. Infrastructure costs money to run.
The same is true for MPP. When an agent pays through MPP on Tempo, the Tempo chain handles settlement natively. But when MPP operates on other chains, like Solana, someone has to run the settlement service. Someone has to hold the fee payer keys, manage the RPC connections, validate transaction safety, and sponsor gas so buyers do not need native tokens.
This is where the protocol war framing breaks down entirely. A settlement operator that supports both protocols serves a strictly larger market than one that supports only x402 or only MPP. The engineering cost of supporting both is incremental, not architectural. The HTTP 402 response format differs. The payment verification logic differs. The underlying settlement infrastructure, the wallets, the RPC connections, the security policies, the chain integrations, all of it is shared.
We operate one of the most active x402 facilitators in the ecosystem, with over 25 million settlements across eight chains. In March 2026 we shipped @dexterai/mpp, a managed Solana settlement service for MPP. Both protocols run on the same facilitator. Same fee payer wallets. Same RPC infrastructure. Same security validation. Same rate limiting. Two protocols, one engine.
A seller using our infrastructure can accept x402 payments and MPP payments with a single account and a single wallet. The buyer does not know or care which protocol the seller prefers. The seller does not know or care which protocol the buyer's agent speaks. The facilitator translates.
What protocols cannot do
Protocols move bits. They do not solve discovery, quality, or trust.
When a research agent needs a satellite imagery API, it faces a problem that no payment protocol addresses. Which provider is reliable? Which one returns useful data? Which one is actually running and not returning stale cached responses? The agent can pay through x402 or MPP with equal ease, but neither protocol tells it where to point.
Two independent research reports published in early 2026, OnchainLu's "Machine Economy 2030" and Dynamic's "Agentic Finance and Stablecoins," both identify discovery as the largest unsolved problem in the agentic commerce stack. When the number of agents and the number of services both grow by orders of magnitude, manually curated tool lists stop working.
Solving discovery requires three capabilities that compound together. Settlement infrastructure that sees real payment data. Quality verification that tests endpoints with real money, not just uptime pings. And reputation signals that are portable and verifiable across platforms rather than locked inside one vendor's database.
We run quality probes against 5,000+ API endpoints every fifteen minutes using real x402 payments. We score them 0 to 100 using frontier AI evaluation. We publish those scores on-chain through ERC-8004 reputation registries on both Base and Solana. Every settlement through our facilitator, x402 or MPP, generates data that feeds this system. The more transactions we process, the better the quality signal becomes. The better the quality signal, the more agents route through our discovery layer. That flywheel does not care which protocol drives the payments.
The Tempo stack is impressive and incomplete
Tempo and Paradigm have built serious technology. This deserves acknowledgment without qualification.
Tempo Transactions are a custom EIP-2718 transaction type that bundles fee sponsorship, parallel execution, batch calls, scheduled payments, and passkey authentication at the protocol level. This is not ERC-4337. It is not EIP-7702, though it evolved from that work at Ithaca. It is a clean native primitive that eliminates the need for bundlers, paymasters, and relay infrastructure that other account abstraction approaches require.
The Accounts SDK, which Georgios Konstantopoulos announced today, brings passkey-backed wallets to any web application with a single Wagmi connector. npm i accounts, add tempoWallet() as a connector, and users authenticate with Face ID or fingerprint. No seed phrases. No browser extensions. The developer experience is genuinely good.
MPP sessions solve a real scaling problem. Per-request on-chain settlement does not work when an agent is streaming LLM tokens at hundreds of requests per second. Sessions make high-frequency micropayments viable by moving payment verification off-chain and batching settlement.
Gondolin, published through Paradigm's earendil-works GitHub organization, provides sandboxed micro-VMs for AI agent code execution with programmable network policies and secret injection that prevents credential exfiltration. Absurd, from the same org, implements durable workflow execution using only Postgres stored procedures, replacing Temporal's 170,000-line Python SDK with under 2,000 lines. These are the internal tools of an engineering team operating at the highest level.
And Tempo has distribution that money buys. $500 million in Series A funding at a $5 billion valuation. Stripe's merchant network. Design partnerships with the largest financial institutions and AI companies in the world.
What Tempo does not have is a marketplace. No quality verification system. No on-chain reputation layer. No agent discovery infrastructure. No protocol-agnostic settlement that works on Solana, Base, Polygon, Arbitrum, Optimism, Avalanche, SKALE, and BSC simultaneously. Tempo has a payments directory with 100+ services listed at launch. That directory is MPP-only and Tempo L1-first.
Tempo built the best possible version of a single-protocol, single-chain payment stack. The thesis of this piece is that single-protocol, single-chain is the wrong architecture for what comes next.
Why protocol agnosticism wins
The history of internet infrastructure is a history of middleware eating protocols.
TCP won over IPX and AppleTalk, but the web was not built by companies that bet on TCP. It was built by companies that operated at higher layers of the stack and treated the transport layer as a commodity. Stripe did not build a credit card network. AWS did not build an operating system. Cloudflare did not design HTTP. They built services on top of protocols that other people maintained and made those services indispensable.
x402 and MPP are both useful, capable, and likely to coexist for years. The Linux Foundation and the IETF are both credible standards bodies. Coinbase and Stripe are both well-capitalized and motivated. Cloudflare has shown through concrete actions that it intends to support both.
The businesses that win in this environment are the ones that sit above the protocol layer. Settlement services that speak both protocols and route payments efficiently. Marketplaces that index resources regardless of which protocol protects them. Discovery systems that evaluate quality based on observed behavior, not protocol metadata. Reputation networks that accumulate trust across chains and standards.
Picking a protocol is a bet on a standards negotiation between companies with more leverage than you have. Operating at the layer above protocols is a bet on execution, which is the only variable you control.
What we are building for
We support x402 across eight chains today. We support MPP on Solana today. When MPP sessions ship in our managed settlement service, we will be the only provider offering session-based micropayments on Solana, a chain that Tempo's L1 cannot reach. When MPP adoption grows on EVM chains, our facilitator already handles those chains for x402 and extending MPP support is incremental work.
Our marketplace indexes over 5,000 resources. It does not care which protocol a resource uses. Our quality verification spends real money testing real endpoints every fifteen minutes. It does not care which protocol the endpoint accepts. Our ERC-8004 reputation scores are published on-chain and accumulate with every settlement. They do not care which protocol generated the settlement.
If x402 wins, we are one of its most active facilitators. If MPP wins, we already speak it. If both protocols coexist, which is the most likely outcome, we are the only settlement provider that handles both natively on the same infrastructure.
The protocol war does not matter. What matters is whether you can settle payments, verify quality, and route agents to the right services at scale, regardless of which standard the next wave of builders adopts. That is the infrastructure we operate, and it works today.
| Facilitator | x402.dexter.cash |
| x402 SDK | @dexterai/x402 |
| MPP SDK | @dexterai/mpp |
| Marketplace | OpenDexter |
| All chains | Supported Networks |