Syrax

Products

Payments Exchange Intelligence Vault Syrax Chain SRX Token Tokenisation Rewards Sentinel

Solutions

All industries

Developers

Syrax Chain for builders Faucet Governance

Company

Insights About Careers Compliance Security Labs Quant
Open the app
Insights · Infrastructure

Institutional Crypto Custody Explained

How institutions safeguard digital assets — and why custody is the foundation of everything else.

Custody — the safeguarding of the cryptographic keys that control digital assets — is the foundation of institutional participation. Get it wrong and nothing else matters.

Why custody is different for crypto

In traditional finance, assets sit in a register held by an intermediary, and a mistake can usually be reversed. On a blockchain, whoever controls the private key controls the asset. A lost key means lost access, and a stolen key means stolen funds, with no central party able to undo it.

Custody in this setting is therefore the management of keys: how they are generated, stored, used and recovered.

The stakes are also higher because of the speed of transfers. A traditional bank transfer can be recalled within a window; a blockchain transaction is final within minutes. Controls therefore have to work before a transfer is made, not after.

It also means that custody cannot be separated from the rest of the operation. The same keys that protect assets also authorise payments, settle trades and sign contract calls, so weaknesses in one area spread quickly to the others.

For that reason, serious custody design treats key management as a core system with its own governance, not as a storage detail.

Models of custody

No single model suits every purpose. Most institutions combine them.

  • Hot storage — keys online for fast access; convenient but higher attack surface.
  • Cold storage — keys offline; highly secure but slower to transact.
  • MPC custody — keys split across parties so there is no single point of compromise, combining security with operational speed.
ModelSpeedExposureTypical use
HotFastOnline, so exposed to remote attackWorking balances for operations
ColdSlowOffline, exposed to physical and process failureLong-term reserves
MPCFastNo complete key in any one placeActive treasury and settlement

The mechanics of the third model are covered in MPC wallets explained.

Many institutions use a tiered approach: a small working balance in a fast, policy-controlled environment, and the bulk held in a slower, better-protected one, with movements between them approved and recorded.

Hardware security modules and dedicated signing devices add another layer in any of these models. They keep key material inside tamper-resistant hardware, so that even a compromised server cannot read it.

What institutions require

Technology is only half of the picture. An institution also has to show that the process around the keys is sound.

  • Robust key management and disaster recovery.
  • Governance controls — approvals, limits and segregation of duties.
  • Auditability and regulatory alignment.

Segregation and recovery

Client assets should be separated from the custodian’s own, so that a failure of the firm does not become a failure of its clients’ holdings. Recovery plans must work when people, devices or sites are lost, and they must be tested rather than assumed.

Regulators set their own expectations on top of this, including how client assets are held, how often they are reconciled, and what is disclosed to clients. Meeting them is part of the design, not an afterthought.

Operational resilience is a further expectation. A custody system should keep working when a data centre fails, a key officer is unavailable or a software dependency has a fault, and it should be tested for that on a regular schedule.

Policy is as important as cryptography

Most custody failures have not come from broken mathematics. They have come from a single employee with too much power, a missing approval step, or a recovery process nobody had rehearsed.

Good custody therefore pairs strong key technology with enforceable policy: transaction limits, allow-lists of approved destinations, time delays on large movements, and approvals from more than one person. Every action should leave a record that an auditor can read.

Insider risk deserves special mention. Approval flows should assume that any single person might be mistaken, coerced or dishonest, and should be built so that one person alone can never complete a sensitive action.

Regular testing closes the loop. Institutions should rehearse recovery, review who has access at set intervals and remove permissions that are no longer needed, since access that is never reviewed tends to accumulate.

Questions to ask of any custody arrangement

Whether an institution builds its own custody or relies on a provider, a consistent set of questions helps compare the options.

  1. Who can authorise a movement of funds, and how many people are needed?
  2. Where are the keys, or key shares, physically and logically held?
  3. What happens if a device, an employee or a whole site is lost?
  4. How are assets segregated from the provider’s own balance sheet?
  5. Can an independent party examine the controls and the records?
  6. What is the insurance position, and what exactly does it cover?

Answers that rely on trust in a single person or a single machine are the ones to treat with care.

Custodial and non-custodial models

The question of who holds the keys splits the market in two. A custodial provider holds them on the customer’s behalf and takes on the safeguarding duty. A non-custodial design leaves the keys with the user, who carries the responsibility for them. Each suits different needs, and the trade-offs are set out in custodial vs non-custodial.

Institutions often need both: a controlled environment for operational flows, and self-controlled storage where they want to keep direct authority.

The decision is rarely permanent. Many organisations move assets between the two as their needs change, for example holding settlement balances with a provider while keeping reserves under their own control.

The Syrax approach

Syrax is designed around that split. The Syrax Payments gateway is built as the custodial side, for institutional flows, using hybrid MPC and hardware-backed security wherever Syrax holds assets. The Syrax Vault is the self-custody side: Syrax never holds or sees a user’s private keys or seed phrases.

Each regulated Syrax service will launch only upon receipt of the applicable licences. The security model is outlined on the security page.

The intent is that each user and institution can choose the model that fits, and that the boundary between the two is always clear: where Syrax holds assets, it carries the safeguarding duty; where it does not, the keys stay with the user.

This article is for information only and does not constitute financial, legal or investment advice. Each regulated Syrax service will launch only upon receipt of the applicable licences.