Institutional treasury custody
Multi-signer hardware-backed cold storage, MPC for hot operations, time-locked withdrawal patterns, and the operational runbook for moving funds under duress.
Blockchain & Web3 · MESH
MPC custody, hardware-backed signing, account abstraction, and the operational discipline institutional custody actually requires. Engineered against the trust model — not the marketing slide.
The problem
Custody failure modes are uniquely unforgiving. A single compromised key, a recovery seed in a desk drawer, a multisig with two signers on the same Yubikey, an MPC vendor whose service degrades under load, an admin pattern that lets one person move treasury — and the loss is permanent. Most custody implementations look secure on the architecture diagram and decompose under operational analysis.
We engineer custody for the actual trust model. MPC threshold signing where the operational benefits earn their complexity, hardware-backed cold storage for treasury, multisig with named human owners and explicit time-locks, and recovery patterns that survive personnel changes. Operational runbooks that assume hostile environments. The custody pattern that ships is the custody pattern that survives an attempted theft, a personnel turnover, and a vendor outage — not just the architecture review.
Where it ships
Specific applications we’ve built and operated. Not speculative — every example below is grounded in a real shipped engagement.
Multi-signer hardware-backed cold storage, MPC for hot operations, time-locked withdrawal patterns, and the operational runbook for moving funds under duress.
Threshold-signature schemes (FROST, GG18/20) for operational hot wallets. Vendor-backed (Fireblocks, Copper) or self-hosted; we tell you which fits.
Account abstraction with social recovery, session keys, sponsored gas, and batched operations. Audited bundlers, paymaster integration, and the operational story for upgrades.
App-side hot wallets for protocol operations (gas tank, oracle bots, automation) with bounded permissions, monitoring, and circuit breakers on anomalous behavior.
Sanctioned-address screening, jurisdictional access controls, audit-trail tooling, and integration with the regulatory-reporting infrastructure institutional custodians require.
How we engage
Each phase has a deliverable, an owner, and an acceptance criterion. Not slogans — operating rules.
Discovery: who can do what, under which constraints, with which recovery path. We design the trust model on a whiteboard before picking a vendor or pattern. The model lands in writing — including the failure modes and the response when each one materializes.
MPC vs multisig vs smart-account custody — picked by fit, not by hype. We model operational complexity, vendor risk, recovery-pattern viability, and regulatory constraints. Some engagements end with 'don't build custody, use Fireblocks / Copper / BitGo' and we say so.
Every custody flow modeled adversarially — what does an attacker do if they compromise one signer? Two? The vendor? Time-lock and circuit-breaker patterns sized to the threat. Test suites against the threat model, not just the happy path.
Monitoring on signing patterns, time-lock states, withdrawal flows. Runbook for incident response, vendor outage, personnel turnover, and the duress signaling pattern. Quarterly tabletop exercises against the runbook.
Capabilities
Stack
Selected work
Common questions
Threat-model dependent. MPC for operational hot wallets where the signing flow needs no on-chain footprint and the trust model can absorb vendor risk. Multisig for treasury where on-chain auditability is critical and the governance group is small. Smart accounts (ERC-4337) for consumer-facing wallets where social recovery and session keys earn the complexity. Most institutional setups use all three at different layers.
Buy, in most cases. The operational discipline required to run custody at institutional grade is hard to achieve from scratch. Build only when (1) the use case is consumer-facing and the vendor solutions are wrong shape, (2) the regulatory frame requires self-hosted, or (3) the operating economics genuinely justify the engineering investment. We tell you which case applies, in writing.
Designed up front. Hardware-backed share distribution with explicit reconstruction quorum, stored in geographically and organizationally distributed locations. Recovery procedure rehearsed quarterly. Personnel-turnover scenarios planned. We will not deploy a custody pattern whose recovery story is 'three people remember it'.
Sanctioned-address screening (Chainalysis, TRM, Elliptic) integrated at the transaction level. Jurisdictional access controls where the regulatory frame requires. Audit-trail tooling that reports to the regulator's expectations, with explicit retention. We've shipped this for institutional clients across multiple regulatory frames.
Yes. EVM-chain custody is straightforward; Bitcoin, Solana, Cosmos, and other non-EVM chains each have their own MPC / multisig story. We've shipped multi-chain custody for institutional clients with the operational tooling that makes the chain differences invisible to the treasury operator.
Trust-model design: 4–6 weeks, $80K–$200K. Operational custody build (vendor-integrated): 3–6 months, $300K–$1M. Self-hosted MPC infrastructure: $1M–$3M. Smart-account / AA wallet platform: $500K–$1.5M. Managed Services: $40K–$150K monthly retainer. External audit and HSM costs pass through. Brackets published honestly so visitors self-qualify before the first call.
Within Blockchain & Web3
Talk to us
A senior engineer plus the MESH department lead joins the first call. No discovery gauntlet, no junior reps.