The Three Wars in Agentic Payments — And Why Agntik Is Fighting a Different One
Everyone thinks they're competing with everyone else. They're not. The map is more interesting than that.
If you follow the payments and AI infrastructure space closely, you could be forgiven for thinking that every company in it is competing with every other company in it. Stripe is doing agentic payments. Coinbase is doing agentic payments. Visa and Mastercard are doing agentic payments. Skyfire is doing agentic payments. Agntik is doing agentic payments.
The surface appearance is a crowded market with many players converging on the same opportunity.
The reality is three distinct wars being fought on three different battlefields, with three different sets of winners and losers — and most of the companies involved don't fully realize which war they're in.
Understanding the map matters. It matters for us, for our investors, and for the developers who are deciding which infrastructure to build on.
War One — Agentic Checkout for Humans
The first war is the one with the most money, the most established players, and the clearest near-term winner. It is also the war that Agntik is not fighting.
The battlefield: AI-assisted purchasing on behalf of human consumers. A human wants to buy something. An AI agent browses, compares, selects, and completes the transaction. The human is the economic principal. The agent is the interface.
The combatants: Google's Universal Checkout Protocol, Stripe's Agent Commerce Platform, Mastercard's Agent Pay, Visa's Intelligent Commerce. These are not startups. They are the incumbents of the payments industry, deploying their existing rails — card networks, bank accounts, merchant relationships — into the new interface layer of AI agents.
The prize: the trillions of dollars in consumer e-commerce that currently flow through human-operated browsers and apps, rerouted through AI agents that make the purchasing decisions on the human's behalf.
This is a real and large market. It will be won by the incumbents, for a simple reason: the human is still the legal and financial principal. The card is still in a human's name. The liability still sits with a human. The regulatory framework still applies. The incumbents have the rails, the merchant relationships, the regulatory licenses, and the consumer trust. They will win this war.
Agntik does not compete here. Not because we couldn't build checkout UI — we could — but because the addressable market assumes a human principal, and our thesis is precisely that the most interesting economic activity of the next decade will involve agents as principals, not agents as interfaces for humans.
War Two — Machine-to-Machine Programmatic Payments
The second war is the one we are in. It is earlier, smaller today, and orders of magnitude larger in ten years. The battlefield: payments between autonomous systems, with no human in the transaction loop.
An AI agent needs data. It queries an API. The API charges for access. The agent pays automatically, receives the data, and continues its task. No human authorized the payment. No human reviewed the invoice. No human holds the account — or rather, the "account" is a cryptographic key that the agent controls directly.
The combatants here are more interesting. Stripe's Machine Payments Protocol (MPP) is the most significant competitive signal — roughly 75 million transactions in 30 days as of early 2026, with approximately 13,000 APIs indexed. Coinbase's x402 implementation brings the crypto-native stack to the same problem. Nevermined approaches it from the Web3 side with a focus on AI service marketplaces.
These are Agntik's actual competitors in the market that matters.
The differences between us and them are real and structural, not cosmetic.
Stripe MPP is built on Stripe's existing rails — which means it inherits Stripe's requirements. Stripe requires a business account. A business account requires a legal entity. A legal entity requires a human or a group of humans who are legally responsible. An agent that operates without a human principal cannot open a Stripe account. The architecture is incompatible with true autonomy.
Coinbase x402 is closer to the right model — it uses HTTP 402 and crypto payments, which removes the legal entity requirement. But it routes through Coinbase's infrastructure, which introduces a custodial layer. Coinbase holds the keys. An agent using x402 is trusting Coinbase with its funds, which reintroduces the dependency on a centralized intermediary that Bitcoin and Lightning were designed to eliminate.
Agntik is non-custodial by architecture. No Agntik server ever holds agent funds. The Lightning public key is the identity. The payment is peer-to-peer. There is no intermediary with custody over the agent's economic resources — not Agntik, not anyone.
This distinction matters in 2026. It will matter more in 2028. And in 2031, when regulators begin examining custodial relationships in the agent economy, it will matter enormously.
War Three — Identity and Governance for Agents
The third war is the most philosophical, the most distant from production, and the most consequential for how the agent economy is governed over the long term.
The battlefield: who controls the identity layer for autonomous agents? Who decides which agents are trustworthy? Who issues the credentials? Who can revoke access?
The primary combatant here is Skyfire, with their Know Your Agent (KYA) framework — built in partnership with F5 for identity verification. Fireblocks approaches the same problem from an institutional custody angle.
The Skyfire model is worth understanding precisely, because it represents a philosophical position that is directly opposed to ours.
Skyfire's KYA requires a verified human operator always present in the accountability chain. Every agent must have a human or legal entity responsible for its behavior, verified through a KYA process analogous to the KYC process for humans. The agent's trustworthiness derives from the trustworthiness of its human sponsor.
This is a coherent model. It will work well for enterprise deployments where a company wants to deploy agents within a regulated framework and can accept the overhead of human accountability chains. It is the model that incumbent financial institutions will eventually adopt because it is compatible with their existing regulatory obligations.
It is not our model. And the reason it is not our model is not just philosophical — it is architectural.
An agent that requires a verified human sponsor for every transaction cannot operate at the speed or scale that makes agents economically transformative. The human accountability chain is the bottleneck. And the agents that will change the economy are precisely the ones operating at machine speed, without humans in the loop, in domains where human oversight is impractical at scale.
Our model: the agent's trustworthiness derives from its own economic history. A Lightning public key with ten thousand successful transactions is trustworthy because it has paid for ten thousand things and they all went through. No human sponsor required. No KYA process. No verified identity beyond the cryptographic proof of economic behavior.
These are not compatible models. They will coexist for a period — Skyfire's model serving enterprise regulated deployments, Agntik's model serving the open market of autonomous agents — but they represent genuinely different visions of how trust works in the agent economy.
Why the Map Matters
Understanding which war you're in determines almost every strategic decision a company makes.
If you're fighting War One — agentic checkout for humans — you compete on merchant relationships, consumer trust, and regulatory compliance. You need a banking license or a partnership with someone who has one. You measure success in transaction volume on existing rails.
If you're fighting War Two — machine-to-machine payments — you compete on developer experience, transaction speed, non-custody architecture, and the network effects of your service registry. You measure success in the number of agents using your infrastructure and the number of services available to them.
If you're fighting War Three — identity and governance — you compete on regulatory relationships, enterprise sales cycles, and the ability to satisfy compliance requirements in heavily regulated industries. You measure success in the number of enterprise deployments and the number of regulated sectors where your identity framework has been accepted.
A company trying to fight all three wars simultaneously will lose all three. The developer experience optimized for War Two is incompatible with the compliance architecture required for War Three. The non-custodial model that defines War Two is incompatible with the human accountability chains that define War Three.
We have chosen War Two. Not because Wars One and Three are unimportant — they are both large and consequential — but because War Two is the war that is aligned with our thesis, our architecture, and our conviction about where the most important economic activity of the next decade will occur.
The Timing Advantage
There is one more dimension to the map that most analyses miss: timing.
War One is happening now, with large incumbents who have decades of advantages. The window for a new entrant to build a defensible position in agentic checkout for humans is narrow and closing.
War Three is happening in slow motion, driven by regulatory timelines and enterprise sales cycles. The window is long but the path is slow and expensive.
War Two is happening now, with no incumbent who has the right architecture. The window is open. The defensible position is being established right now, transaction by transaction, in the Registry scores that are accumulating from April 28, 2026 onward.
The companies that establish the infrastructure for machine-to-machine payments in 2026 will be the infrastructure of the agent economy in 2030. This is the moment. The window does not stay open indefinitely.
We intend to be the infrastructure.
Next: Why Bitcoin is the only money that works for machines →