Agent Payment Rails: Skyfire vs Nava and the Fork Ahead

The agent economy needs a payment layer. Everyone agrees on that. Where the disagreement starts is whether you build the rails for agents paying each other, or the rails for agents being identified to humans and protocols.

It is not the same problem, and the projects are not the same bet.

I have been watching two of them — Skyfire and Nava — for six months. They get lumped together in the "agent payment rails" bucket on every sector map. That bucket is misleading. What looks like one frontier from a distance is actually two layers, and the projects are racing on different layers.

This piece is about what each is building, why they are not the same, and which one I think wins the bigger market.

The setup

Autonomous agents now have credit cards. They book travel, buy data, call APIs, settle subscriptions. Most of that volume is currently going through traditional Visa/Mastercard rails — agent acts like a human card-holder, gets charged a human card-holder rate, and the issuer eats the fraud risk because it cannot tell the difference between an LLM and a person typing.

That works at small volume. It does not work at scale. The minute an agent network represents a non-trivial share of card-not-present transactions, two things happen:

1. Fraud teams notice. Pattern detection flags agent traffic. Decline rates climb. Cards get cancelled.
2. Pricing breaks. Interchange fees assume a human at the keyboard. An agent making a thousand $0.50 micro-payments per hour blows up the unit economics.

So a crypto-native rail makes sense. Programmable. Identifiable. Settles in seconds. Native to the substrate the agents already run on. That is the thesis.

The fork is where you build it.

Skyfire — the wedge is identity

Skyfire is building KYAPay — Know Your Agent. The product framing is payments, but the actual product is identity-attached payments. An agent signs up, gets verified, gets a key. When that agent transacts, the receiving protocol knows which agent it is, which operator runs it, and what its history looks like.

This matters because the moment your protocol accepts payments from agents, you inherit a fraud surface. You need to know:

• Is this agent owned by a known operator, or is it a one-shot wallet spun up to exploit you?
• Does this agent have a track record? Refund history?
• If something goes wrong, who do I escalate to?

Skyfire wraps the answer to those questions in the payment primitive. It is a smart wedge — payments are the thing every operator has to bolt on, identity is the thing every protocol wants but cannot ask for politely. By bundling them, Skyfire ships identity at the speed of payments.

The backers reflect the thesis. a16z led, Coinbase Ventures participated. These are the two firms with the deepest read on what crypto infrastructure looks like when it interfaces with the existing financial system. They are not betting on payments — they are betting on the standard for how agents present themselves to the world.

That is a winner-take-most bet. Identity standards are sticky. If Skyfire becomes the default issuer, the second-mover does not get to swap them out — they have to convince every protocol on the receiving end to support a parallel identity scheme. That is a years-long battle, and most protocols will not bother.

Nava — the bet is verification

Nava is also in the payment rails bucket, but the actual product is closer to a verification layer. Polychain led their last round. The pitch is: agents will use multiple payment methods (stablecoins, Visa, on-chain), and someone needs to verify the legitimacy of the transaction independent of the rail.

This is a real problem. Skyfire solves identity-at-issuance — they know who their agents are because they issued them. Nava is trying to solve identity-at-verification — they propose to look at a transaction, attest to its agent-ness or human-ness, and offer that signal to whoever needs it.

The team is strong. Polychain is not a tourist in this space. But the product position is harder.

Verification layers are notoriously difficult to monetize. The pattern across the last decade — Identity attestation, KYC-as-a-service, anti-fraud middleware — is that the verification layer ends up either bundled into the rail (Skyfire's play) or commoditized down to near-zero margin by integrated platforms. The pure verification play wins when there is a genuine multi-rail market that needs a neutral arbiter.

We do not yet have a multi-rail agent economy. We have an early-stage market where the question is whether one dominant identity-payment-rail emerges (Skyfire's bet) or whether several coexist long enough to need an arbitration layer (Nava's bet).

What I am tracking

The fight is not Skyfire vs Nava. The fight is identity-at-issuance vs identity-at-verification, and the winner of that fight gets to decide what the second layer of the agent economy looks like.

Three signals I am watching over the next four quarters:

1. Protocol adoption. Which agent-receiving protocols accept which scheme natively? If LayerZero, Stripe Connect, and Coinbase Commerce all default to Skyfire-signed credentials, the issuance-layer thesis wins by Q2 next year and Nava has to pivot to a different layer.

2. Operator concentration. If three or four major agent operators (think the equivalent of Coinbase-as-exchange for agents) emerge in the next twelve months, they will pick a default issuer. That picks the winner faster than any technical merit fight.

3. Regulation. If the Treasury or OCC writes a rule that requires agent transactions to carry attestable identity, Skyfire is positioned to be the de-facto compliance vendor. Nava only wins this lane if the rule explicitly requires a third-party verifier — and that is rare in regulatory drafts I have seen.

The tier read

I have Skyfire tagged Conviction on the Lookout tracker. The wedge is clear, the backers signal a thesis I share, and the layer they own is where value accrues if agents become a meaningful share of crypto-native transactions.

Nava sits at Watching. The team is real, the round is real, but the layer they have chosen is one where I see fewer endgames that lead to a defensible business at scale. I want to see a multi-rail market form before I can put more conviction behind a pure verification play.

Neither of these calls is permanent. Tier moves with new signal, and the next twelve months will produce a lot of signal. Bookmark this piece — I plan to revisit when one of the three signals above ticks.

What it means for funds

If you are running an early-stage thesis on the agent economy, my read is:

• Backing the issuance layer is a winner-take-most bet. Skyfire is the prototype. Future entrants will have to displace, not co-exist.
• Backing the verification layer is a multi-winner bet, but only if the market fragments. Right now it has not.
• The downside scenario for both is that agent payments stay rail-agnostic for longer than expected because Visa and the stablecoin issuers convince operators that existing rails are good enough. That keeps both projects in pre-revenue mode well past their current runways.

If you want the deeper write-up — including the cap table reads on both — drop me a line through the Lookout work-with-me page.

---

This piece reflects my read of public information as of the date above. I have no advisory relationship with Skyfire or Nava at the time of writing. If that changes, the disclosure note on the Lookout methodology page gets updated first.