Core Vaults for Stablecoins
Idle stablecoins quietly cost money. Chasing yield costs more. The answer is a product that puts capital to work inside limits it can actually defend.
The moment a company starts using stablecoin rails at real volume, balances start to pile up. Settlement float, idle deposits, operating reserves, customer money waiting to move – it accumulates whether anyone planned for it or not.
And what’s left sitting there is a problem with two bad answers. Leave it idle and you’re burning the yield it could be earning. Reach for whatever APY looks best and you’ve taken on risk nobody running a real balance sheet could defend.
Neither is a product. That gap is the whole reason every wallet, exchange, and fintech suddenly wants an Earn button – and the reason the hard part isn’t the button, but what a stablecoin is even allowed to do once it lands.
Yield needs a reason now
It comes down to two regulatory moves, both drawing the same line from different sides.
GENIUS is already law and working through implementation. It stops permitted payment stablecoin issuers from paying holders interest or yield solely for holding, using, or retaining a payment stablecoin – no passive return just for keeping the coin in a wallet.
CLARITY, the market-structure bill still moving through the Senate, would push that logic past the issuers. Its Section 404 would bar covered digital asset service providers and their affiliates from paying the same kind of passive, deposit-like yield on stablecoin balances.
What neither one touches is yield that comes from actually doing something. That is the line being drawn: not against yield, but against yield that accrues to a balance for sitting still. To earn now, a dollar has to be put to work inside a product, with real strategy and real risk behind the return. And building that product is a different job entirely from minting a coin.
So what is the product?
On Mellow, that product is a Core Vault.
Think of it less as a place stablecoins go to earn a number and more as a mandate: a set of rules, written straight into the contract, governing what a given pool of capital can do – which assets it holds, where it can route, how much goes where, who is even allowed to deposit. Whatever the approved strategies earn inside those rules is the yield.
What sits inside depends entirely on the mandate, and the spread is wide. A conservative book might hold tokenized treasuries and money-market funds and nothing else – BlackRock’s BUIDL, Ondo’s USDY, Superstate’s USTB, Franklin Templeton’s BENJI, each carrying its own issuer eligibility and jurisdiction rules. A more active one layers in onchain lending through approved markets like Aave, Morpho, or Compound, then stablecoin AMMs and short-duration positions that double as the vault’s liquidity buffer, and basis trades that reach centralized venues through Copper ClearLoop and Ceffu without the vault ever giving up custody.
A single vault can run all of that at once, inside whatever the mandate approves. The accounting is what holds it together: multi-asset bookkeeping and an oracle-defined NAV collapse a spread of positions, each with its own price source, settlement clock, and liquidity profile, into one coherent number. Adding a venue later is usually a config change rather than new contract code.
There is no fixed menu here, only settings on a dial. A regulated platform that needs defensible underlying assets turns it one way; a fintech with more appetite turns it the other; the machinery underneath doesn’t change.
And it isn’t only consumer Earn. A company sitting on a stablecoin float can point the same architecture at its own balance sheet – hold this much liquid, put the rest to work within approved limits, never cross them. There the end user is the company itself.
The thing institutions care about
The number on the front of an Earn product is the easy thing to show, and it is not what a serious partner is buying. The moment a vault touches a payment company’s float or a neobank’s deposits, the question quietly shifts from what’s the APY to what is this thing actually allowed to do, and who decides. A clever strategy with no hard boundary around it is just a liability with good marketing.
Core Vaults enforce more than sixty onchain permissions at the contract level: allocation caps per strategy and per venue, asset and venue whitelists, oracle checks that run before anything executes. The pieces doing the enforcing are called verifiers, and every action gets checked against the mandate before it happens. The model is composable, so a routine rebalance might clear through a single verifier while a high-impact move needs several plus a time lock. Guardrails can be updated through structured governance, but they cannot be breached mid-execution. Step outside the mandate and you can’t make a transaction. The curator gets to run the strategy; the curator does not get to run off with it.
This is also what lets an autonomous agent operate a vault without it being reckless. The verifiers don’t care whether the thing calling a function is a person, a multisig, or a model – only whether the call sits inside the mandate. An agent can rebalance and manage liquidity at a cadence no human treasurer would keep up with, and still has no way to act outside its limits, because the code won’t allow it. Whether a model can make the decision was never really the hard question; whether you can let it act on that decision safely is.
Compliance lives at this same layer, written in rather than bolted on. The vault carries eligibility hooks for KYC and KYB providers, transfer agents, accreditation registries, and jurisdiction rules, plus per-asset and per-issuer handling for regulated instruments that come with their own transfer restrictions. A platform in one country and a platform in another can run entirely different compliance perimeters on the same infrastructure, with no forked codebase.
The part that quietly breaks
Here is the failure almost nobody prices in until it happens. The yield doesn’t run on crypto time. Many tokenized treasuries settle on a T+0 to T+2 cycle, some positions only exit on a set window, and meanwhile the person who deposited expects a clean APY on screen and a withdrawal that clears the moment they tap the button. Those things are in tension, and a product that pretends otherwise ends up with soft accounting and exits that quietly reward whoever leaves first.
Handling that gap is engineering. Core Vaults run deposits and redemptions through asynchronous queues, and track economic NAV and withdrawable NAV as genuinely separate figures, so the settlement reality stays inside the machinery instead of leaking onto the user, who just sees something that works. It is unglamorous, and it is the piece you only learn to respect after watching a simpler design come apart on a heavy redemption day.
The stablecoin finally has a job
Stablecoin yield is moving off the coin, where it is no longer welcome, into the product layer, where the strategy, the mandate, and the controls live around the dollar instead of inside it.
That layer is quiet and technical and easy to walk past, which is precisely why it is the part that decides who wins. It is where a company’s float, a fund’s reserves, and before long an agent’s treasury stop being money that sits and start being money that works, inside rules anyone can read.
Today, putting a stablecoin to work is the optional and often build-it-yourself thing. Soon it becomes the floor and the thing a platform is assumed to have before anyone takes it seriously.
The full technical reference – architecture, strategy allocation, risk enforcement, and compliance – is here: