Curve Finance

Curve started from Michael Egorov's StableSwap whitepaper in late 2019 and launched in early 2020 (Curve News). A normal AMM prices every pair on a curve that spreads liquidity across all possible prices. StableSwap packs most of it around a 1:1 ratio, so a large USDC-to-USDT swap moves the price very little, as long as the two assets stay near peg.

The trade-off is how a pool behaves when an asset loses its peg. A StableSwap pool keeps swapping at close to 1:1 until its balance becomes lopsided. LPs in a depegging pool end up holding mostly the failing asset. That is useful to know before providing liquidity for a small stablecoin because of its rewards. Curve's own score card is here.

crvUSD is Curve's dollar stablecoin, minted by borrowers against collateral. Its liquidation engine, LLAMMA, is different from Aave's or Compound's. As collateral falls into a band, it is gradually converted to crvUSD, and converted back if the price recovers. A borrower loses some value on the way down instead of taking a single liquidation hit, and can still be fully liquidated if the price keeps falling.

PegKeepers hold the price near $1. When crvUSD is scarce in its stable pools, they add supply. When it is plentiful, they withdraw it. In August 2026 a burst of demand, much of it linked to YieldBasis, pushed PegKeeper debt from $33.75M to $108.2M in six days. The mint rate fell to 1.55%, and borrower-minted crvUSD nearly doubled to $72.2M (Curve News). Hourly prices stayed within about 6 basis points of $1 that month.

LlamaLend markets are separate lending pools where anyone can lend crvUSD against one collateral asset. V2, live on Ethereum in 2026, adds yield-bearing collateral such as sreUSD and svZCHF. These markets are permissionless, and that has a cost: after the 10 October 2025 crash, several LlamaLend markets on volatile collateral took bad debt. The CRV-long market left lenders with impaired withdrawals for months (Curve News).

Curve's governance was switched on in an unusual way. On 13 August 2020 an anonymous user, 0xc4ad, deployed the CRV and DAO contracts from code Curve had published for review. Curve verified the deployment and the community adopted it as the real launch. The model those contracts introduced, vote-escrow, is now copied across DeFi. You lock CRV to get veCRV, which carries voting power and a share of protocol fees, and the longer the lock, the more veCRV you get.

veCRV votes also decide which pools receive CRV emissions. That set off the "Curve Wars", where protocols such as Convex accumulated veCRV or paid its holders to direct rewards to their own pools. Over five years the DAO handled 1,276 proposals (Curve News).

• Core development: funded by the DAO through grants to Swiss Stake AG, which reports quarterly.
• Risk: LlamaRisk has provided risk oversight; in August 2026 the DAO selected yRisk as its risk provider, with funding executed on 2 September.
• Treasury: since June 2025, 10% of protocol revenue goes to a DAO-controlled reserve. Before that, almost all fees went to veCRV holders and LPs.
• Emergency DAO: a multisig that can cut, but never raise, crvUSD debt ceilings and pause specific components. Its powers were widened in November 2025.

Curve's core swap maths has held up well. Its losses came from the layers around it: the compiler, the website domain and the newer lending markets.

The 2023 incident is the lesson that stuck. Curve's contracts were correct as written, but the Vyper compiler versions used for some pools produced a reentrancy lock that did not work. Whitehats, including the bot 0xc0ffeebabe, saved about $17M while the attack was still running (Rekt). After the second DNS hijack, Curve moved its front end to curve.finance and has encouraged users to reach it through ENS rather than a registrar domain (Curve News).

CRV token data is on our market page.