Oracles

Merum needs an accurate, manipulation-resistant price for LBTC to decide how much a borrower can borrow and when a position must be liquidated. A wrong price is one of the most damaging failure modes for any lending protocol: a price that is too high lets borrowers over-borrow, and a price that is too low triggers unfair liquidations.

Merum addresses this with a composite oracle that combines three independent sources rather than trusting any single one.

Price sources

• Pyth — a low-latency, pull-based oracle network publishing prices aggregated from many institutional contributors. Provides a fresh, high-frequency price.
• Chainlink — a widely used push-based oracle with a decentralized network of node operators. Provides a robust, independently-sourced reference price.
• On-chain TWAP — a time-weighted average price drawn from on-chain liquidity. A TWAP is expensive to manipulate because an attacker would have to move the price and hold it moved across the averaging window, not just for a single block.

How they combine

The oracle router reconciles the three feeds before a price is used for any borrowing or liquidation decision. The design goals are:

• Agreement check. Feeds are compared against one another. If they diverge beyond a configured tolerance, the price is treated as unreliable.
• Freshness check. Each feed must have updated recently. A stale feed is rejected.
• Manipulation resistance. The TWAP component blunts attempts to push a single instantaneous price up or down to trigger profitable liquidations or borrows.

If the sources disagree too much or a feed is stale or unavailable, the protocol can pause price-sensitive actions for the affected market rather than act on a price it cannot trust. This favors safety (temporarily blocking borrows and liquidations) over acting on a bad number.

Why this matters to you

• As a borrower, the multi-source design reduces the chance that a brief, manipulated price spike on one venue causes an unjustified liquidation of your position.
• As a supplier, it reduces the chance that a manipulated low price lets a borrower over-borrow and leave the pool with bad debt.

No oracle is perfect. A correlated failure or a successful manipulation of multiple sources at once remains a residual risk, disclosed in the risk disclosures. Exact feed identifiers, tolerances, and staleness windows are published under Risk parameters at launch.