What Stripe's investor letter says about Nuri

What Stripe actually wrote

The letter is worth reading in full. Here is the short version.

Stripe announced OpenRouter as its largest acquisition ever. The companies did not disclose the price. The New York Times reports $7.5 billion; Reuters reports slightly more than $8 billion. OpenRouter had been valued at $1.3 billion only three months earlier.

Before OpenRouter came Bridge, Privy and Metronome. Bridge cost $1.1 billion. Metronome's official terms were not disclosed, but reporting puts the deal at about $1 billion. Privy's purchase price was also undisclosed; its last reported valuation before the acquisition was $230 million. That gives Stripe roughly $9.6 to $10.1 billion in reported deal value across OpenRouter, Bridge and Metronome, plus whatever it paid for Privy. It does not include the cost of building the rest of the stack.

They wrote: "We decided that January 1st marked the beginning of the singularity, and we have since been operating on that basis." They report H1 net revenue up 41% year over year, free cash flow up 43%, 88% of the Forbes AI 50 on Stripe, and token consumption on OpenRouter compounding at 9% per week.

The core sentence is this one: "building economic infrastructure for the internet is mostly the same thing as building the economic infrastructure for AI."

And they name the pieces they are building for it: Stripe Projects and Directory for agent onboarding and discovery, Metronome for usage, Bridge and an Agentic Commerce Suite for payment, Tempo as their own blockchain for agents, MPP as a machine payments protocol, Privy for wallets, and Open Standard, their own stablecoin.

They also wrote something I keep coming back to: "There is a fear that AI will yield unemployment or centralization; perhaps both. We think that it is important that deployment of AI enhances human agency."

I agree with every word of the diagnosis. I disagree with the medicine.

The part where they describe our product

Go through Stripe's list of primitives and put it next to Nuri.

Stripe now owns Privy. Privy calls its user wallets non-custodial and offers private-key export, which is a real escape hatch. But before export, signing still depends on Privy infrastructure: one encrypted key share is stored by Privy and the other can only be decrypted and combined inside its AWS Nitro Enclave. If that infrastructure disappears before the user exports, the wallet does not have a local recovery path. With Nuri, WebAuthn PRF derives the user's own key locally from the passkey. In multisig wallet models this is the user key, not every signer key. The co-signing path is time-limited through CSV, so after a fixed block window the user retains an independent recovery path. A separate backup key, such as a hardware wallet, can add a 2-of-3 recovery option. The same principle can be applied to EVM Safe accounts used by agents. The user does not have to remember to press an export button before Nuri disappears. That difference is not cosmetic. It is independence built into the wallet.

Nuri brings fiat and crypto into one wallet, but pointed at people and their agents rather than only platforms: card and IBAN, bitcoin onchain and over Lightning, stablecoins, and the open rails that connect them. An agent may operate a user-owned wallet within limits, or own a wallet and act autonomously. Both modes matter.

The important question is not where a budget rule happens to be enforced in one implementation. It is whether the authority is real. An agent needs enough freedom to act, spend, earn and complete work without asking a human about every cent. The owner still needs a way to define the boundaries and revoke delegated access.

Nuri is AI-native below the interface. The system is operable through CLI and MCP, while onchain actions remain client-side where the wallet model requires it. The frontend can be the web app, the iOS app, the user's own agent or the Nuri agent on WhatsApp. Every important capability has to work without assuming a person is clicking through screens.

We are not building a chain and we are not issuing money. Nuri is designed to be agnostic across chains, protocols and financial services, including the fiat side. Open source and open standards matter because the user should never depend on one company, including ours, to keep using their money.

Same singularity, different answer

Here is the honest way to say it: we believe what Stripe believes. Agents will hold money, work inside budgets, pay for services, subscribe, settle and earn. Stablecoins will carry a lot of it, but they are not the whole future. Fiat, bitcoin and Lightning are part of the same picture.

Stripe's answer is vertical integration: intelligence routing, billing, wallets, a chain, a stablecoin and a directory. It is an impressive machine and it will work for a lot of businesses. Privy makes this more nuanced than a simple custodial-versus-non-custodial argument; it supports non-custodial and exportable wallets too. The concern is concentration. More and more of the agent economy can begin, move and settle inside one commercial universe. Human agency, as long as Stripe agrees.

Nuri's answer is to compose rather than own. Self-custody where the user chooses it. Autonomous ownership where the agent needs it. Open source where independence depends on it. Open standards so another service can replace us without replacing the user's financial life. Chain-agnostic and service-agnostic, including on the fiat side.

This is not a head-to-head fight with Stripe. They serve platforms building for agents at enormous scale. We start with the person and the agent working for them, or with an autonomous agent that needs a real financial identity of its own. The thesis is the same. The architecture and incentives are not.

Agents with an actual bank account

This is the part I am most excited about. An agent should be able to own a wallet and operate autonomously. It should also be able to act through a person's wallet with delegated authority. Those are different relationships, and Nuri has to support both.

When the wallet belongs to the user, the passkey and keys stay with the user. The agent receives only the authority it needs. When a wallet belongs to an autonomous agent, the agent must be able to hold money, earn, spend and settle without waiting for a human click every time. Autonomy that stops at the payment screen is not autonomy.

The fiat side is what turns this from a crypto demo into a financial product. Nuri connects agents to cards, IBAN and bank transfers as well as bitcoin, Lightning and stablecoins. An agent may need to pay a supplier by bank transfer, keep a card funded, settle over Lightning or pay an API with an open machine-payment protocol. These are not competing worlds. They are the world money already lives in.

The architecture is chain-agnostic and service-agnostic. Capabilities are exposed through MCP and CLI rather than locked inside one frontend. The interface can be our web or iOS app, the user's own agent, or the Nuri agent on WhatsApp. The backend should not care which surface made the request, only whether that request has the authority to act.

Verification follows need, not ideology

One design decision confuses people at first and then clicks: Nuri serves both sides, retail and institutions, and verification follows need.

You want a self-custodial wallet for bitcoin and stablecoins, no account, no forms? That exists, that is live, and it works worldwide. You want the full picture with card, IBAN, and banking rails? Then KYC applies, because that is what the regulated side requires, and our partners handle it properly. A business wants the same thing with KYB and controls? Same framework, same tools, different verification tier.

This is not fence-sitting. It is the only architecture that can actually be worldwide. Most of the planet cannot pass European KYC and should not need to just to hold their own money. And European users who want an IBAN should not have to leave the product to get one. One wallet, one passkey, and the user decides how deep into the regulated world they want to go. Agents inherit exactly the tier their human has.

Five people and a machine

Stripe has thousands of employees. We are five.

That is only possible because Nuri is AI-native all the way down. The backend is built to be operated through CLI and MCP. Onchain actions happen client-side where ownership requires it. The frontend is interchangeable: web, iOS, the user's own agent, or the Nuri agent on WhatsApp. A graphical app is one interface, not the product boundary.

Support runs through an agent cockpit, with humans handling the cases that need a human. Content, product videos, monitoring and distribution increasingly run through agent pipelines. The website is not only readable by agents; it is operable by them. That is why agent readiness matters to us more than a design award.

I will not pretend this is all smooth. Building a company this way means debugging your own future every day. But every claim we make about the agent economy is one we test on ourselves first. We are our own first customer.

Where this goes

Stripe's letter ends with gratitude to their investors and a promise of intensity. Fair. Here is ours.

The wallet keeps growing as the bridge between fiat and crypto for people, and becomes the same bridge for their agents. Bitcoin, Lightning, stablecoins and banking services should feel like parts of one financial life, not separate products owned by separate gatekeepers.

The biggest company in payments just declared that the economic infrastructure of the internet and the economic infrastructure of AI are becoming the same thing. Cloudflare is building identity, wallets and payment infrastructure around its network. Circle is building autonomous wallets around stablecoins. PayBox is connecting agents to existing wallets, exchanges and cards. These are strong signals that the market is real.

The risk is not that these companies are building. They should build. The risk is that identity, intelligence, wallets, payment rails and settlement collapse into a few vertically integrated clouds. Open source and open standards are how the rest of us keep the system composable. Self-custody and exportability are how users remain independent of the companies serving them.

One direction we want to pursue is a collaboration with Nous Research around Hermes. This is a target, not an announced partnership. Hermes is open source, runs locally or in persistent cloud environments, and can connect to hundreds of models through different providers or custom endpoints. It already supports provider routing, fallbacks, tools, MCP and messaging channels. Put that together with Nuri and the idea becomes simple: agents on click. Choose a model, launch an agent locally or in the cloud, give it a wallet or delegated authority, and let it operate across the Nuri universe without locking the user into one model, one cloud or one financial provider.

That is the future we want to build toward: direct access to many LLMs, routing across models, durable cloud agents and financial tools in one open-source, non-custodial system. The agent may belong to a person or stand on its own. The surface may be an app, WhatsApp or no graphical interface at all. What matters is that intelligence and money can meet without either becoming a new point of captivity.

Nuri does not need to own every layer. It needs to connect the layers without becoming the reason a user is trapped inside them. That is a different kind of ambition: not to become the financial operating system that nobody can leave, but to build one that still works when they do.