How to vet a vault curator before you allocate

How do you check a curator's record before you allocate to one of its vaults? Among the questions Morpho's curator explainer suggests are "How have their strategies performed across different market cycles, including stress periods?" and "How much capital do they currently curate across all their vaults?" The listing in Morpho's app answers neither well. Four steps answer them from the chain:

1. List every vault the curator has run, not the ones shown today. The 13 curators with five or more vaults on Ethereum have run 185; 41 are listed today. Of the unlisted vaults that still report more than $100,000, most are losses or unpaid interest.
2. Read each vault's result in real assets, not in its share price. 16 vaults show at least $1,000 in Morpho's API with under 1% of it actually in their markets, and their share prices have not moved.
3. Treat assets that keep growing in a fully borrowed market as unpaid interest. One unlisted vault reports $902.7M. All of it is supply in a market where nothing has been repaid since November 2025.
4. Look past the setup checklist to what the vault lends against. On 28 February 2026 the 16 largest USDC vaults looked almost the same on roles and timelocks. They differed on how their collateral was priced.

As of 6 November 2026, for MetaMorpho vaults on Ethereum. Figures are read from the chain and from Morpho's API on that date.

1. Every vault, not the listed ones

A curator sets a vault's strategy: which markets it lends into and how much. Morpho's listing policy calls listing "a discovery decision", not "a risk assessment". A vault listed for more than three months that holds under $100,000 "may be delisted at any time", and "where all of a curator's vaults have been delisted, the curator is no longer displayed". New listings must now be Vaults V2, so many older vaults leave the app as curators move on. Morpho's GraphQL API still returns every vault, with listed: false on the hidden ones.

There are 470 MetaMorpho vaults on Ethereum, and only 137 carry a curator name in that API. Grouping the rest by the addresses Morpho lists for each curator gives this, for every curator with five or more:

Most unlisted vaults are empty, small, or were never meant for outside money. The ones that matter still hold something. Across all 470, 32 unlisted vaults report more than $100,000, $1.44B in total. Nineteen of them, $1.40B, either booked a loss or show a share price more than ten times where it started. The eight in the second group are all stablecoin vaults, where a jump like that only comes from interest piling up in a fully borrowed market (step 3). Gauntlet USDC Frontier, Resolv USDC and Clearstar Yield USDC are among them; Gauntlet USDC Core, which took its March loss in its share price, is still listed.

Names move too. Two Steakhouse vaults were renamed and handed to Waterline in June 2026, so their history under Steakhouse now sits under another name. "Alpha USDC Core", delisted in August, is called "Hawkwood" today.

2. Real assets, not the share price

How a loss shows depends on the vault's contract version. On version 1.0, Morpho's docs say, "the share price of the vault will instantaneously decrease". On version 1.1 "the share price will not decrease", the loss is booked as lostAssets, and "the last vault users will face the bad debt". The booking also waits for the vault's next transaction, so a vault can hold nothing and show no loss yet.

So read real assets per share: the vault's supply in each of its markets, added up, divided by its shares. Of the 382 version 1.1 vaults, 16 report at least $1,000 in Morpho's API, $9.5M together, while under 1% of it sits in their markets. Fifteen have booked the gap as lostAssets; Resolv USDC, at $1.03M, holds nothing in any market and has booked nothing yet. Gauntlet USDC Frontier booked 1.91M and still shows a share price of 1.08. Its depositors and Gauntlet USDC Core's were later paid about half their per-share loss through Merkl claims, which no share price shows.

One of the 16 is still listed. SingularV ETH carries only a low-liquidity warning, and the $17,079 the API shows as its size is the 7.6 WETH it booked as lost in June, about 0.7% of the vault then. Depositors who left afterwards were paid in full; the three holders still in own a claim on nothing.

3. Interest received, not interest booked

When a Morpho market is fully borrowed and nobody repays, its rate keeps rising and the interest is added to suppliers anyway, as covered in how to monitor a curated lending vault. The market against Elixir's staked deUSD held 8.86M USDC, all borrowed, in November 2025. At its last update, on 29 October 2026, it showed $14.6B supplied and $14.6B borrowed.

Adpend USDC, a vault with no curator name, holds a share of that market and nothing else: $757.7M at that update. Morpho's API adds the interest since then and reports $902.7M, with a share price of 791. It is not listed, but a table sorted by assets would put it near the top. Morpho's second question, capital curated across all vaults, needs the same check: how much of it could anyone withdraw.

4. The collateral, not the setup checklist

The usual checks are roles, timelocks and a guardian. The owner can replace the curator; a timelock delays some changes, such as raising a cap or removing a market; a guardian can revoke those while they wait. Fee changes, a new curator or a new allocator take effect at once. On 28 February 2026, three weeks before the USR exploit, the 16 MetaMorpho USDC vaults above $10M held $1.01B between them:

• every timelock was 72 or 168 hours;
• 15 of 16 had a guardian;
• 14 of 16 were owned by a multisig.

What differed was the collateral. Nine of the 16 had money in markets whose oracle reads no market price: an issuer's reported rate, a vault exchange rate or a Pendle formula. That was $360M of the $1.01B. Every vault with USR exposure on 22 March was among the nine, since the USR oracles are issuer feeds, though two of them held under 1%. Five of the nine held none, so this is not a warning sign on its own. It tells you what you are underwriting: in those markets a loss shows up when the issuer's number changes, not when a market price does.

The same snapshot shows who else can move your money. Ten of the 16 let Morpho's Public Allocator move their supply into markets at anyone's request, up to limits the vault sets. On 22 March that took one vault from 0.01% to 20.56% exposed in a day, as the monitoring piece shows.

What to ask the curator

• Every vault you have run on every chain, including closed and delisted ones, with dates and addresses.
• Each vault's result in real assets per share, and any loss booked as lostAssets.
• Fee changes over time, and anything paid to depositors outside the vault, such as compensation through Merkl, with who was eligible.
• For each market: what its oracle reads, and the Public Allocator limits into it.
• Your own capital in your vaults, and who holds the rest of the shares.

Then check each answer against the chain.

Limits

• Operators are grouped by address. A curator's vaults run from addresses Morpho does not list for it are counted as unnamed.
• "Real assets" is Morpho's own accounting. Supply in a market whose collateral the oracle still prices at par counts in full until the loss is booked.
• This covers MetaMorpho vaults on Ethereum. The same curators run vaults on other chains and on Vaults V2, and a Vault V2 that allocates into a version 1.1 vault does not show that vault's losses in its own share price.
• A record is not a forecast. It shows how a curator's vaults fared, not how they will.

Doing this for any curator

One chain and one vault type took reads across 470 vaults, their roles, names, markets and oracles. A real diligence file covers every chain and protocol a curator touches, with the vaults that closed. Datablocks indexes every EVM chain from genesis, keeps failed tokens and protocols, and can show for any manager or vault every position they've held, through every market. If you allocate to curated vaults, book a call below and tell us which curators you need to check.