Non-custodial design

What it means

Locked liquidity sits in a per-asset, per-chain AssetVault (each chain's aqUSDC, for example, is a separate contract), not a team-controlled wallet, and the team cannot move locked assets to an arbitrary address through ordinary contract calls. Locking liquidity does not mint a tradeable share token, it credits your principal instead, a scalar claim denominated in the underlying asset.

Your principal backs the vault's strategies by consent, not by a token transfer, recorded as a non-transferable, read-only position rather than something you can sell or move to someone else. Capital is either idle (sitting in the vault's shared pool) or deployed (currently sourced out to a strategy you've consented to back, working through a registered adapter such as JIT liquidity or a cross-chain strategy). Realized gains and fees accrue to a separate withdrawable balance, credited in proportion to your actual participation in that strategy, and there's no share price to track.

Sync versus async withdrawal

• Free principal (idle capital) withdraws instantly, with no request/wait step.
• Deployed principal withdraws asynchronously, first requesting then claiming once the protocol frees the capital (adapter unwind / settlement). This is the flow used when your principal is currently backing a strategy rather than sitting idle.

See Async redeem semantics for the full request/claim lifecycle if you're integrating against it.

The never-trapped guarantee

An async redemption request becomes claimable when either of the following happens.

1. the capital is attested free (the protocol's signer confirms the position has been unwound/settled), or
2. a timeout backstop elapses, regardless of attestation.

Every timeout is bounded by an immutable, compile-time ceiling of 7 days that no owner, governance, or upgrade can raise. So even if attestation fails entirely, every request becomes claimable on its own within 7 days, with no discretionary step required. Each route's timeout is configured independently and can never be tuned past that 7-day ceiling. The claim itself pays out once the freed capital is physically back in the vault, and a claim that arrives before the capital does reverts cleanly and can be retried.

What non-custodial does not mean

It does not eliminate the following risks.

• smart contract risk (the vault, adapters, and Composer are new contracts under audit, see Audits & security process)
• chain risk
• bridge/messaging risk in cross-chain flows
• off-chain trust in narrowly-scoped signing keys (see below), where the team can't arbitrarily redirect funds, but this isn't fully trustless, since a small number of bounded, monitored roles have real (though limited) power.

Bounded, monitored roles

Vault access control is role-based, each role a distinct key held by a distinct signer, and every role is scoped narrowly enough that no single one can drain the vault.

• No key can move locked assets to an arbitrary address, since capital only moves through registered, allow-listed venues.
• Pausing and resuming are separate powers, so a single compromised key can freeze the vault but can't also unfreeze and drain it, since resuming requires a separate, slower authorization.
• Upgrading the vault implementation passes through a timelock delay, giving a response window before any change takes effect.

The hybrid PnL model, the one place off-chain trust remains

Your withdrawable balance is not fully self-attested. Same-chain realized PnL is settlement-derived and trustless, applying straight from the settlement event whenever an adapter settles a position, and no key can move this number.

Cross-chain in-flight PnL works differently, since capital deployed on another chain, before the bridge/message finalizes, can't be observed on-chain in real time.

At beta's settings, a report can move a strategy's valuation by at most 25 bps per day. The rate is tunable by governance, but every setting is clamped by an immutable per-epoch ceiling that no owner action can raise. The clamp bounds the rate of change, not a lifetime total, and every report is reconciled against the real settlement amount well before drift could add up to anything material. A compromised reporting key can therefore nudge a valuation by a bounded amount per epoch. It cannot drain the vault, move funds, or hide a real loss. That is a materially narrower trust assumption than a withdrawal-authorizing role, and it exists only to cover the in-flight cross-chain gap that same-chain settlement can't observe on its own.