Before we begin, we'd like to address the Kelp hack over the weekend. It triggered a major liquidity crisis on Aave, with nearly $10 billion frozen. How did Morpho weather the episode?
There was zero contagion on Morpho, and that's exactly the point. We built an infrastructure where every deployment is isolated. There were exactly two markets listing Kelp's token (rsETH) on Morpho — perfectly siloed — representing about one million dollars in total borrowings. Against a TVL of $12 billion, that's negligible. Take Coinbase: during their first six months, they were only exposed to Bitcoin; now they've added Ether. When rsETH blows up, it changes absolutely nothing for them. It's the exact opposite of a protocol like Aave, where a single failing collateral asset contaminates the entire pool. On Aave right now, the collateral is rehypothecated and illiquid, which blocks liquidations and, by extension, affects the USDC and USDT markets. That's precisely why every fintech is coming to Morpho: they want to control their own liquidity and risk, not depend on a list of 35 collateral assets where one problem can take down the whole protocol.
Could the situation still spiral?
Honestly, yes. The real question is where Kelp will realize the loss. Either the holders on mainnet absorb the shock, or you make the users who bridged to Layer 2s pay — which would be extremely harsh for every Aave deployment on Optimism, Mantle, and Arbitrum. On both sides, you've got an army of lawyers. There are people publicly claiming the loss will be realized on the L2s, even though no official decision has been made. These are legal negotiations where everyone is staking out a position to rally as much support as possible. And even if we told people to exit Aave, they can't get out.
Let's turn to your latest news. You recently unveiled Morpho Agents, designed to make it easier for AI agents to operate on your protocol. What's your vision for the agentic economy, and where does Morpho fit in?
This might sound a bit philosophical, but I believe we're heading toward two types of computing. Deterministic computing and probabilistic computing. The first will be carried by blockchains and smart contracts — code execution where you need absolute certainty about the outcome. Everything else will be probabilistic execution, driven by AI. You're not 100% sure of the outcome, but it's so much more flexible and adaptable that it'll be enough. The two complement each other extremely well. It's more or less obvious to everyone that an agent needs funds to act quickly. But the moment you give it money, you have to constrain it — you need guarantees on what it's doing. So there's a dual challenge: how does the agent navigate financial rails, and how do you ensure that its execution — probabilistic by nature — stays within acceptable bounds?
At Morpho, we're focused on this vertical. We're staffing it up significantly. We expect the market share of agents interacting with the chain to be very large. We're giving them, first, a deep understanding of the protocol and how to interact with it — to generate yield or take out a loan. We've also launched a kit for developers building on Morpho. If you distinguish between a "user agent," which uses the protocol, and a "builder agent," which creates applications that run on Morpho, the builder agent dramatically cuts integration time.
How many agents are actually executing transactions on-chain today?
The market isn't mature yet. People throw around figures like 130,000 AI agents having registered an on-chain identity since January, but those aren't meaningful metrics. With a single server, I could spin up two million agents creating wallets from my PC. The only metric that will matter is dollar volume. Today, it's all testing. We don't have a single serious user running an agent to invest in a real vault. The builder agent, on the other hand, is seeing heavy usage and strong growth. It's a bet, but the direction is pretty clear. Look at ChatGPT's integration with Starbucks: you connect your card, the agent can order your coffee. I find it hard to tell myself that's not the future and that we shouldn't be investing in it.
"The traditional financial system is absolutely not designed for agents"
What does blockchain actually bring to the agentic economy that traditional financial rails can't?
It all comes down to programmability and guardrails. Having tested banking payment APIs myself, I can tell you the traditional financial system is absolutely not designed for agents. A blockchain is native territory for an agent: you have an open context, all the data is accessible, you create an account with a wallet in two seconds, you can transact in fractions of cents using stablecoins, you program precise limits — a spending cap of X dollars per day, a whitelist of authorized smart contracts. That level of precision and programmability simply doesn't exist in traditional finance.
The contrast is actually quite striking: people in AI are a thousand times more bullish on crypto than crypto people themselves. They all see the potential to revolutionize financial rails through autonomous agents, while in crypto right now, we're a bit weighed down by hacks and crises.
The fund A16z (also an investor in Morpho, editor's note) recently floated the idea of storing agent-related data on-chain to continuously verify their permissions and provenance. Is on-chain identity for agents a relevant concept in your view?
Absolutely. And I'll go further: identity in general is something sorely missing from the on-chain economy. For us at Morpho, it would be extremely useful because it serves as a trust signal. As a lending protocol, our job is to match lenders with borrowers. Today, the only way for a borrower to demonstrate reliability is to show their collateral. But that's not the only trust signal out there — traditional finance knows this well. It would be so much better to tie a verified identity to an on-chain profile, along with a repayment history and a credit score. That would allow for far more precise underwriting.
For AI agents specifically, even though they're not human, you can record a lot of relevant information on-chain: the capabilities assigned to them through their smart account, the amount of available funds, limits, the protocols they're allowed to interact with. In some cases, it's also useful to link the agent to a human identity.
Are you thinking of World (formerly Worldcoin) in particular?
Yes, I really like what World is doing. They have an application, for instance, that lets you take out an under-collateralized loan — around $1,000, with nothing down — just using proof of personhood. The mechanism is simple: if you don't repay, they'll never lend to you again, and your attestation is unique. If you have an agent in World, you can tag it with your identity. You can't game the system by multiplying loans. The agent is tied to you, you're accountable for its actions, and it impacts your credit score. It's a primitive that's going to be fundamental. And by the way, World is starting to find a business model: they integrated with Tinder two days ago — you can swipe more if you're verified as human. So in a sense, Tinder is already using blockchain.
If an agent executes a bad transaction — or even something illegal — who's responsible?
I'm not a lawyer, but if I had to think it through, I'd start from the purpose of liability: being able to hold someone accountable and sanction an inappropriate action. There are several schools of thought. Some say it's the agent's developer. Others go all the way back to the GPU manufacturer — there are voices in the United States arguing that Nvidia shouldn't export chips to China, which effectively places responsibility at the hardware level. That's excessive, in my opinion. The more reasonable position is that responsibility falls on the person who launched the agent. AI still has no consciousness or will of its own. There's an initial prompt, a launch action, and it's that initial intent that should be called into question — the same way that when you run code to hack someone, it's the person who presses the button who's identified as responsible.
"Tempo pivoted all-in on agentic at breakneck speed"
Among institutional players, have you already been approached about Morpho Agents?
We're getting a lot of inbound, especially from distributors and applications building fairly advanced services. I can't name names, but what I can say is that every blockchain is working on this. Tempo pivoted all-in on agentic at breakneck speed — it's a major bet for them. Coinbase is doing interesting things on their end. Everyone is starting to build their own agentic layer.
We're seeing a battle of standards emerge: x402 from Coinbase, AP2 from Google, MPP from Tempo, and so on. Where do you stand?
Everyone's pushing their own standard, trying to form consortiums, and then some newcomer proposes a standard to unify all the standards, and someone else pushes back because theirs wasn't picked. It's always kind of the same story. We're not in that fight. We'll be agnostic: we integrate everything, every standard, and we'll see which one the market converges on. Besides, it's not obvious to me that the agentic world even needs a single standard. In a deterministic world with code execution, having multiple standards is a headache. But in an agentic world where the agent is inherently adaptable, fragmentation is less of an issue.
Are you already deployed on Tempo?
We were deployed before the public launch. Generally speaking, Tempo excites me. I'm a big fan of Stripe, which is behind it. It's a strong team, coming out of Paradigm, with phenomenal core devs who came from Ethereum. It's also, paradoxically, a pro-Ethereum signal.
"Agents will accelerate winners and kill off protocols resting on their laurels faster"
Coinbase is a major partner for Morpho. Is Morpho Agents also designed with these kinds of large B2B partners in mind?
Yes. There are big B2B partners who've expressed interest. They're building their own services for their users where agents will be assigned to each client, and those agents will need to use lending services around them. But where we're seeing the most interest is the builder agent. The use case is clear: you're a developer at a large fintech, you want to integrate Morpho fast, cut your time-to-integrate from six months to two — you use the Morpho Builder Agent.
If agents end up generating most of the on-chain volume tomorrow, does that shift how value is distributed across protocols?
I struggle to see why it would change anything fundamental. What it will do, though, is accelerate competition. Agents are more rational: they'll arbitrage much faster. Today, there's enormous inertia in DeFi. Users stick with an underperforming protocol for years because it's been around a long time, it's big, and — up until this weekend, anyway — people trusted it. When you're an agent, you don't have that inertia. You migrate immediately to the protocol offering the best risk-return profile. My gut feeling is that agents will accelerate winners and kill off protocols resting on their laurels much faster.
Morpho was just officially recognized as a unicorn by French Tech. Is that important to you?
It's nice of the government to have taken the time for this recognition, because it wasn't straightforward for them. There's no equity in the traditional sense — it's a token. So it's a somewhat unusual unicorn, a bit atypical. Personally, I spend very little time in the traditional startup ecosystem. But the reality is, if you want to reach the general public, if you want credibility with banks and asset managers, if you want to help shape sound regulations for DeFi, you need a baseline of institutional credibility. We want to show that we're not actors who hide. We want to be able to engage with the real world. It's a necessary step if tomorrow, on Morpho, we aim to host billions in loans and borrowings, and capture a slice of the $200 trillion global credit market.
>> DeFi: Apollo makes strategic investment in Morpho
>> Morpho's profile on The Big Whale
%201.png)
%201.png)
%201.png)
%201.png)
%201.png)
