Cowrie Research

llamma-primary-sources-2026-05-14

4 min read

LLAMMA primary-source snapshot

Fetched: 2026-05-14 UTC

1) crvUSD overview

Source: https://docs.curve.finance/developer/crvusd/overview

Selected excerpts:

crvUSD is designed to provide a more capital-efficient stablecoin mechanism and smoother liquidations, while maintaining a decentralized design which the Curve DAO governs.

LLAMMA is the market-making contract that rebalances the collateral. As the name suggests, this contract is responsible for liquidating collateral.

The crvUSD mint markets use several iterations of the Controller and AMM (LLAMMA) contracts … Llamalend later adopted the V3 blueprint.

Additional signals from the source:

  • This page is the clearest proof that LLAMMA should be treated as a reusable lower-layer engine, not just a one-off crvUSD implementation detail.
  • The overview also makes governance and market-admission power visible: new collaterals are added through DAO approval and markets are instantiated through factory/blueprint tooling.
  • The shared V3 blueprint with Llamalend is especially important for corpus organization.

2) LLAMMA explainer

Source: https://docs.curve.finance/developer/crvusd/llamma-explainer

Selected excerpts:

LLAMMA (Lending Liquidating Automated Market Maker Algorithm) is a market-making algorithm that rebalances the collateral of a loan using an AMM.

Unlike other liquidation mechanisms that have a single liquidation price at which the loan is liquidated once reached, LLAMMA has a liquidation range in which collateral is continuously being rebalanced.

Liquidation is not done on a user basis but rather on a band basis, making the liquidation mechanism scalable.

Additional signals from the source:

  • This page best explains the core conceptual move: liquidation becomes continuous inventory rebalancing across bands instead of a one-shot cliff event.
  • The band-based liquidation path is the main reusable insight because it changes who shares execution conditions and how liquidation scales.
  • The soft-liquidation / de-liquidation split also exposes where losses and recoveries actually occur.

3) LLAMMA contract docs

Source: https://docs.curve.finance/developer/crvusd/amm

Selected excerpts:

LLAMMA … is responsible for liquidating and de-liquidating collateral based on market conditions through arbitrage traders.

The AMM (LLAMMA) contract is shared across all Controller versions. Unlike the Controller, the AMM API has remained stable — the same contract is used by both crvUSD mint markets and Llamalend markets.

Collateral is put into bands by calling deposit_range() … Values for n1 and n2 are already determined in the Controller contract.

Additional signals from the source:

  • This page shows that the arbitrage path is not incidental; traders are the mechanism that performs soft-liquidation and de-liquidation.
  • It also makes the Controller-versus-AMM split clear: the Controller owns loan policy while the AMM owns collateral-rebalancing execution.
  • That separation is an important comparison point versus lending systems where liquidations are more monolithic.

4) Canonical AMM implementation

Source: https://raw.githubusercontent.com/curvefi/curve-stablecoin/master/curve_stablecoin/AMM.vy

Selected excerpts:

@title LLAMMA - crvUSD AMM

  • p_oracle - external oracle price (can be from another AMM)
  • active_band - current band. Other bands are either in one or other coin, but not both
  • bands_x[n], bands_y[n] - amounts of coin x or y deposited in band n

@notice LLAMMA constructor … @param _price_oracle External price oracle which has price() and price_w() methods

Additional signals from the source:

  • The raw contract confirms that oracle price, band accounting, and dynamic-fee logic are foundational state, not just documentation vocabulary.
  • It also confirms that LLAMMA is built as a dedicated AMM with explicit collateral-token / borrowed-token roles and controller-admin assumptions.
  • The contract-level glossary is useful because it compresses the real mechanism into comparison-ready variables.

Notes

  • LLAMMA cleared the bar because it exposes a distinct liquidation primitive beneath Curve’s broader stablecoin and lending branding.
  • The most reusable insight is that liquidation can be modeled as oracle-guided continuous rebalancing inside a band grid, with hard liquidation reserved for exhausted health.
  • Main control surfaces to compare later: band selection, controller-admin loan policy, oracle dependence, arbitrage efficiency, fee tuning, and the switch from soft to hard liquidation.

Graph View