Scored category

Bitcoin custody providers and control models compared

This page helps you compare Bitcoin custody and control models before you open every provider site. Every provider here is scored on Pledge’s v3.2 8-factor custody methodology, so you can weigh control models and Pledge scores side by side and open the profile that fits.

Coverage refreshed through October 10, 2026. All 11 providers are scored, each launch score carrying a verified or estimated-confidence flag that we revise as deeper verification completes.

Providers in scope
11
The current set of custody profiles under active review.
Coverage state
All scored
Every provider carries a v3.2 8-factor Pledge score.
Score posture
8-factor scored
Each score carries a verified or estimated-confidence flag.

Compare the current custody set before you open seven tabs

This comparison layer separates provider control models and shows each one’s Pledge score, fit, recovery, and service boundary side by side.

Scored8.2/10

AnchorWatch

Control model

Trident Vault is a miniscript "multisig of multisigs" — your own keyset (a single hardware key, or a client-side 2-of-3) paired with AnchorWatch's independent institutional keyset, and both sides must complete their own signatures before coins move. AnchorWatch is a required co-signer while the vault is insured.

Recovery angle

A lost key is recoverable via AnchorWatch co-signing, with Lloyd's-syndicated insurance covering named theft and key-loss scenarios on top.

Fit

Best for users who want collaborative custody plus explicit policy language, not just wallet architecture.

Slow down on

Insurance should widen your questions, not end them. Coverage scope and claims assumptions still matter.

Read profile
Scored8.1/10

Unchained

Control model

Standard client-controlled vault is 2-of-3: you hold 2 keys and Unchained holds 1. Unchained cannot move the bitcoin without one of your keys.

Recovery angle

If Unchained's site is unavailable, the wallet config file plus your two keys recovers the vault in Caravan or Sparrow. If you still hold one key, Unchained can co-sign a recovery.

Fit

Best for users who want guided self-custody rather than a fully outsourced custody relationship.

Slow down on

Read the exact service boundary and recovery assumptions carefully before treating it as “set and forget.”

Read profile
Scored8.0/10

Anchorage Digital

Control model

Single-custodian model under a federal national trust bank charter rather than a state trust.

Recovery angle

Recovery and continuity are institutional process questions under federal supervision.

Fit

Best as a benchmark for federally regulated qualified custody, not as a Bitcoin-native self-custody alternative.

Slow down on

A federal charter does not by itself resolve concentration or product-fit questions for individual holders.

Read profile
Scored8.0/10

BitGo

Control model

Pure institutional-custody benchmark with less user-participation than collaborative models.

Recovery angle

Recovery and continuity are likely service-process questions, not self-custody questions.

Fit

Best used as an institutional reference point, not as a default recommendation for every Pledge reader.

Slow down on

A large brand and broad platform are not the same thing as segment fit or legal clarity for your specific use case.

Read profile
Scored7.8/10

Fidelity Digital Assets

Control model

Institutional-custody benchmark with traditional-finance trust signals rather than user participation.

Recovery angle

Recovery and continuity are likely enterprise-service questions, not wallet-design questions.

Fit

Best for institutional benchmarking, not as a blanket default for self-custody-minded users.

Slow down on

Brand trust and category fit are different things. Users should not confuse the two.

Read profile
Scored7.8/10

Coinbase Custody

Control model

Single-custodian qualified custody — Coinbase Custody Trust Company holds and signs.

Recovery angle

Recovery and continuity are enterprise-process questions, not user-held key design.

Fit

Best for institutional and ETF-adjacent contexts, less obviously the right answer for individual holders who want structural redundancy.

Slow down on

Coinbase Custody Trust and the Coinbase consumer platform are separate legal entities — do not conflate the protections.

Read profile
Scored7.7/10

Onramp

Control model

Individual and IRA vaults are 2-of-3 across Onramp, BitGo Trust, and CoinCover. CoinCover holds a backup key for recovery if an institution fails, is compromised, or a client loses access. The client holds no keys. Business accounts use Onramp, Tetra Trust, and CoinCover.

Recovery angle

If Onramp is unavailable, the other two institutions still have the 2-of-3 quorum and coordinate recovery into a replacement vault. Inheritance planning is included.

Fit

Best for users who want the structural benefits of multisig without operating the keys themselves.

Slow down on

Independence of the three signers is the load-bearing claim — confirm it before treating the model as airtight.

Read profile
Scored7.6/10

NYDIG

Control model

Single-custodian NY-trust custody, Bitcoin-only by design.

Recovery angle

Recovery and continuity are institutional processes, not user-held key designs.

Fit

Best for Bitcoin-focused institutional and wealth-advisory contexts, less obviously a direct answer for individual self-custody-minded holders.

Slow down on

Bitcoin-only framing is meaningful but does not substitute for a clear current read on disclosures and concentration risk.

Read profile
Scored7.5/10

Casa

Control model

Standard is a 2-of-3 vault. You hold your own keys, and Casa holds one Casa Key that co-signs for recovery. Higher plans add a 5-key or 6-key vault.

Recovery angle

Recovery and inheritance are central. If Casa shuts down, you recover with your own keys in Sparrow or Electrum.

Fit

Best for users who want durable self-custody with guidance, not passive institutional storage.

Slow down on

Users still inherit personal operational responsibility, so complexity tolerance matters a lot.

Read profile
Scored7.3/10

Gemini Custody

Control model

Single-custodian NY-trust custody operated by an exchange-linked entity.

Recovery angle

Recovery is enterprise-process, not user-held key design.

Fit

Best for institutional and benchmarking contexts; readers should keep custody and exchange products clearly separated.

Slow down on

Compliance reputation is real, but does not erase concentration risk or the importance of separating the custody product from adjacent Gemini services.

Read profile
Scored6.8/10

Swan

Control model

Swan Vault is 2-of-3: you hold 2 keys in cold storage and manage the third, a Cloud Key, through Swan. Swan has access to only that one key. Swan Safe is separate: a regulated custodian holds the keys.

Recovery angle

If Swan is unavailable, download the Swan Vault Recovery Kit and import it into Specter Desktop or Bitcoin Core, then spend with your two Jade devices.

Fit

Best for users who want Bitcoin-native relationship management more than pure software autonomy.

Slow down on

Do not flatten “concierge” into “safer” until the control model and contingency plan are directly reviewed.

Read profile

Who this provider comparison helps most

This page is strongest when you are still figuring out which control model deserves the next click. It is weaker if you already know your provider shortlist and only want a final ranked recommendation.
Who this is for

Good fit

Use this path when these signals match.
  • Readers choosing between collaborative custody, self-custody support, and institutional custody.
  • Users who want a plain-English map of which provider models fit, alongside each provider’s Pledge score.
  • People trying to settle control-model questions before they compare loans or stablecoins.
Who should pass

Wrong fit

Switch paths when these signals match.
  • Users who already know the exact provider they want and only need the detail page.
  • Readers expecting a mature best-to-worst ranking today instead of an early-coverage compare layer.
  • Anyone whose real next step is loan pricing or stablecoin structure rather than provider/control-model selection.
Scored8.2/10

AnchorWatch

Insurance-backed collaborative custody

Interesting because the insurance structure changes the trust conversation versus plain multisig or third-party custody. The policy story is part of the product, not just a marketing add-on.

AnchorWatch changes the custody conversation by making insurance scope part of the product thesis.

What we'd still verify

The binding policy wording — full exclusions list and claims process — against the public policy summaries, plus how the operating record develops beyond its short tenure.

Scored8.1/10

Unchained

Collaborative multisig

A natural bridge from the current loan research because collateral control, key distribution, and private-client service quality already matter to many Pledge readers.

Collaborative multisig with private-client support, for holders who want to keep a key.

What we'd still verify

Recovery paths, service boundaries, and customer-agreement detail at the provider level — the score is graded from public materials while that deeper pass completes.

Scored8.0/10

Anchorage Digital

Federally chartered qualified custody

The only fully operational federally chartered crypto bank in the US (Anchorage Digital Bank, N.A.), offering qualified custody across 100+ assets — Bitcoin alongside Ethereum, stablecoins, and other tokens — plus trading, settlement, staking, and tokenization services for institutional clients and certain high-net-worth individuals.

Anchorage Digital is the only US crypto custodian fully operating under a federal bank charter — a structural differentiator worth tracking carefully.

What we'd still verify

Customer-agreement scope and the most current regulatory posture — the federal trust-bank charter, SOC-2 audit, and HSM-based custody design are publicly documented.

Scored8.0/10

BitGo

Qualified institutional custody

An important benchmark for institutional custody because it anchors the “qualified custodian” end of the category and gives the research a clear incumbent reference point.

BitGo is less about guided self-custody and more about institutional infrastructure, permissions, and scale.

What we'd still verify

Current contractual terms and service-boundary detail at the institutional tier — the trust charter, SOC-2 audit, and named insurance facts are publicly documented; the service-scope read still deepens.

Scored7.8/10

Fidelity Digital Assets

Institutional custody

A meaningful brand benchmark for the category even if access, disclosures, and retail relevance differ sharply from Bitcoin-native operators.

Fidelity Digital Assets is valuable in custody research as an incumbent benchmark, not as a direct substitute for every Bitcoin-native custody product.

What we'd still verify

A stronger provider-level evidence set on access, service scope, and current disclosures — incumbent institutional custody publishes less product-level detail than Bitcoin-native operators.

Scored7.8/10

Coinbase Custody

Qualified custody (NY trust)

The most heavily used qualified-custody provider in US crypto, operated under Coinbase Custody Trust Company, a New York state-chartered trust.

Coinbase Custody is the incumbent qualified-custody benchmark in the US — useful in research even where it is not the right answer for individual holders.

What we'd still verify

A clean direct read on segregation mechanics, insurance scope, and the most recent disclosures — the legal structure itself is among the most-documented in US crypto.

Scored7.7/10

Onramp

Multi-institution custody (2-of-3)

A Bitcoin-focused multi-institution custody product where signing is split 2-of-3 across Onramp, BitGo, and CoinCover. Designed to remove single-point-of-failure risk without making the user run hardware. The product also supports USD and USDC cash balances alongside BTC.

Onramp is one of the clearest Bitcoin-native answers to the "I don't want to manage keys, but I also don't want one custodian" problem.

What we'd still verify

Legal segregation, insurance scope, and recovery flow at each of the three key signers (Onramp, BitGo, CoinCover) — the score grades the legible structure while that evidence pass completes.

Scored7.6/10

NYDIG

Qualified custody (NY trust)

A Bitcoin-focused institutional platform — founded in 2017 by Ross Stevens (also founder of Stone Ridge) — that combines regulated custody and execution with a separate bitcoin mining and power-infrastructure business. Trust-company custody operates under New York trust framing.

NYDIG is a Bitcoin-focused institutional platform that pairs regulated custody and execution with power and mining infrastructure — narrower in scope than multi-asset custodians and routinely used inside advisor and wealth channels.

What we'd still verify

Legal segregation, insurance posture, and how the institutional product is being offered today — disclosures have shifted in recent years and the score is graded from current public materials.

Scored7.5/10

Casa

Self-custody coordination

A strong fit for the self-custody audience. The key question is not rate shopping but recovery design, inheritance readiness, and how much support a user wants without surrendering control.

Casa is best understood as a self-custody coordination layer, not as a conventional custodian.

What we'd still verify

Recovery execution, membership-service boundaries, and current public disclosures — the score grades the legible architecture while that tighter read completes.

Scored7.3/10

Gemini Custody

Qualified custody (NY trust)

A New York limited-purpose trust company operated by Gemini (Gemini Trust Company, LLC), holding 100+ cryptocurrencies with SOC 2 Type 2 + ISO 27001 attestations, $125M digital-asset insurance ($100M cold + $25M hot, as of Mar 2024), and offline HSM-based key storage in geographically distributed facilities. Integrates with the consumer Gemini Exchange dashboard but is a separate fiduciary entity.

Gemini Custody is a NY-trust qualified custodian with a long compliance track record (SOC 2 Type 2, ISO 27001, $125M insurance — $100M cold + $25M hot, Mar 2024) — useful as a category reference, with the usual caveats about exchange-linked custodians.

What we'd still verify

A current evidence pass that clearly separates the trust-custodian product from adjacent exchange and earn products — the compliance posture itself is among the longest-documented in US crypto.

Scored6.8/10

Swan

Collaborative multisig (Swan Vault, 2-of-3) + segregated trust custody (Swan Safe)

Relevant because many users deciding where to borrow also need a higher-trust answer for where to hold BTC in the first place, especially when they want guided service rather than DIY coordination.

Swan sits closer to private-client Bitcoin concierge than to a pure vault-only custody tool.

What we'd still verify

The exact control, recovery, and provider-boundary details behind the concierge service — the score reflects the high-level model while that evidence review deepens.

Control-model comparison FAQs

This page sets expectations clearly: strong intent, real provider coverage, and early-stage scoring discipline.
Question

Is this a ranked “best provider” list?

Yes. Every provider here is scored on Pledge’s v3.2 8-factor custody methodology and sorted by Pledge score. Each launch score carries a verified or estimated-confidence flag, openly revised as deeper verification completes.

Question

What does the confidence flag on each score mean?

Every provider is scored, but a launch score is marked verified (entity and core facts confirmed) or estimated (some sub-factors inferred from public disclosure). Pledge publishes the score now and tightens the confidence as the control model, recovery path, service boundary, and source set are verified in more depth.

Question

Why does Pledge include both collaborative and institutional providers here?

Because they are all part of the real control-model decision set, even though they solve different jobs. The page keeps them together for search intent and comparison, then separates their tradeoffs so each provider’s Pledge score is read alongside its control-model context, not in place of it.

Control-model updates

Follow control-model updates from the hub once this page narrowed the shortlist.

This comparison page should help you separate provider models, not ask for another dedicated signup. Keep control-model updates centralized on the hub so custody coverage has one clear follow path.

If control models are no longer the real decision, switch paths

Many readers land here while deciding whether to fix storage first, borrow against BTC, or move into a Bitcoin-linked dollar. Keep those alternatives explicit so control-model research routes people honestly.
Deepest coverage

Explore loans

Use loan research when the control-model baseline is already good enough and the real question is lender terms, APR, liquidation policy, and borrower fit.

See loan research
Adjacent category

Explore Digital Credit

Use Stablecoins research when the next question is where to park dollar value in a Bitcoin-linked system, not who should control the Bitcoin itself.

See Stablecoins research
Disclosure policy

Custody coverage is scored — and the disclosure standard came first

Custody coverage exists because key control, recovery, and legal segregation deserve their own scored research. It is not monetized today, and any future provider referrals or sponsorships will be disclosed separately from loan relationships.

No provider ranking is being shaped by an active referral program today.

If Pledge later adds provider referrals or sponsorships here, those relationships will be disclosed separately from loan programs and labeled on the relevant control-model pages.

Future control-model compensation should identify whether it is per-lead, sponsorship, or another arrangement instead of hiding under a generic affiliate label.

Any provider listed without a full ranked score is waiting on the evidence threshold — never on a commercial consideration.

Rankings, methodology, and evidence standards stay independent either way. Read the full disclosure policy.