A multisig Bitcoin wallet is one that needs more than one key to spend, so that losing any single key, or having any single key stolen, does not empty it. Whether you need one depends on how much the coins matter to you, how many years you plan to leave them alone, and whether anyone besides you will ever have to get at them.
Here is the short answer. If your Bitcoin is a small slice of your net worth and you are the only person who will touch it, a single hardware wallet with a well-stored seed phrase is enough for most people. If the amount would change your retirement, or it is shared with a spouse or a business, or you are planning for heirs, a 2-of-3 multisig is worth the extra work. The rest of this page explains how it works, what it costs in money and effort, and how to tell which side of that line you are on.
The question usually comes up right after the first hardware wallet arrives. You write down 24 words, put them somewhere safe, and then realize that those 24 words are now the whole account. Anyone who finds them can take everything, and if they burn, so does your savings. Multisig is the standard answer to that single point of failure. It is not the only one, and it comes with costs of its own.
How a 2-of-3 wallet decides who can spend
Multisig wallets are described as "M-of-N": N keys exist, and any M of them must sign before coins can move. In a 2-of-3, three separate keys are created, usually on three separate devices, and any two can approve a payment. Trezor's explainer calls 2-of-3 the standard setup, secure without being overly complex, and 3-of-5 (five keys, three needed) the higher-security option for larger holdings.
The closest everyday comparison is a bank safe deposit box that needs both your key and the bank's key. Multisig extends that so any two of three keys will do. That one change is what matters. Lose one key and the remaining two still open the wallet. Have one stolen and the thief holds a key that opens nothing on its own.
The parts you have to keep
A single-signature wallet can be rebuilt from its seed phrase alone. A multisig wallet cannot. Each device holds its own seed, and the wallet itself is defined by a combination of all three devices' extended public keys (xpubs), which are the public halves that let software compute every address the wallet will ever use, plus the signing rule and the derivation path. That combined record is called the wallet descriptor or configuration file.
This is the part most people miss. Trezor warns that losing even one xpub can stop you from sending funds even if you still have enough private keys. ForkLog's explainer puts it the same way: without the configuration file, coins may become inaccessible even with the required number of keys. So a 2-of-3 has five things to back up, not one: three seeds, and the descriptor stored alongside them.
What a transaction looks like
The software that builds transactions is called the coordinator. It usually runs on a computer and holds only public information, so it can watch the balance and draft payments but cannot sign them. It produces a partially signed Bitcoin transaction (PSBT), a draft that each signing device adds its signature to. Once two devices have signed, the coordinator broadcasts it.
Multisig addresses look different, too. Per Trezor, legacy multisig addresses start with "3", while native SegWit multisig (P2WSH) addresses start with "bc1" and are longer than ordinary single-signature addresses.
What it protects against, and what you give up
Unchained's Tom Honzik calls the benefit "threshold security": no funds are lost if any one key, device or component is destroyed, misplaced or stolen. In practice that covers three kinds of failure.
- Loss. A house fire takes one seed backup. The other two keys still sign, and you move the coins to a new wallet.
- Theft or coercion. Someone who finds one seed, or one device, cannot spend.
- A flawed vendor. If one manufacturer's devices turn out to generate weak keys, a wallet that mixes manufacturers still has a key the attacker cannot reach. This stopped being theoretical in July 2026, as the next section shows.
- No single seed, device or location can empty the wallet
- One lost key is recoverable in a 2-of-3; up to two in a 3-of-5
- A flaw in one hardware maker's devices is contained if the other keys come from other makers
- A key can be handed to an executor or trustee without giving that person the power to spend
- Three devices and five backups (three seeds plus the descriptor) instead of one device and one seed
- Spending is slower when signers are kept in different places, by design
- Transactions carry more script and signature data, so network fees are higher than single-sig
- The setup cannot be changed in place: moving from 2-of-3 to 3-of-5 means building a new wallet and sending the coins to it
- More steps means more chances for human error; Trezor advises individuals against going beyond five keys
Two of those costs need a closer look.
Fees. Trezor says multisig transactions cost more because they carry more data, and that fees can rise sharply when the network is congested, so it suggests transacting in quiet periods. ForkLog says Taproot removes this drawback. Both can be true: Taproot can make a multisig spend look like a single-signature one, but only if your wallet software and devices build Taproot multisig, and the step-by-step guides cited on this page use P2WSH, which is the older and larger kind. For a long-term holder who spends a few times in a decade, the fee difference matters less than for someone moving coins often.
Rotating a key. If one key is lost or suspect, you do not swap it out. You build a new wallet with a fresh key and send everything to it. Trezor notes that this is an on-chain transaction, and that it can merge your separate coin deposits (UTXOs) into one, which reveals to anyone watching the chain that they belonged to the same person.
The Coldcard flaw: a real test of multisig
On July 31, 2026, Coinkite disclosed that its Coldcard hardware wallets (the Mk3, Mk5 and Q, with the Mk3 worst affected) had been generating seeds with far less randomness than intended. Casa's write-up says the flaw dated to 2021: firmware skipped the device's hardware randomness chip and used a predictable formula instead. Mk3 units on firmware 4.0.1 to 5.0.3 had about 40 bits of effective entropy against an intended 128, which Casa says an attacker can search in hours. Coinkite put the newer models at about 72 bits; independent analysis estimated a worst case near 32 bits for a targeted attacker.
The seeds looked normal. Nobody needed the physical device to take the coins. Fixed firmware shipped the same day, but it only protected seeds made afterwards.
According to ForkLog's report on Galaxy Research's figures, Galaxy's broader estimate including suspected cases reached 2,417.35 BTC (roughly $153 million), and it contacted 190 victims. Every confirmed loss came from a single-signature wallet.
None of the confirmed thefts came from addresses protected by multisignature.
That does not mean every multisig wallet was safe. The detail is in how the keys were made. Unchained's advisory broke it down for its 2-of-3 vaults: if one of the three keys came from a Coldcard, the vault stayed protected, because one cracked key cannot meet a threshold of two. If two of the three came from Coldcards, the funds were at risk. Nunchuk's incident report named multisig quorums built entirely from Coldcard keys as the most exposed group after single-sig wallets, since an attacker with enough cracked keys could work out the wallet's addresses and sweep it. Nunchuk reported that none of its subscribers lost coins, and that about 97% of funds in all-Coldcard quorums had been moved by August 11, 2026.
One more detail mattered: address reuse. Spending from an address publishes its public key on the chain. Nunchuk explains that for affected keys, those exposed public keys let attackers match against pre-cracked seeds, while addresses never spent from stayed protected as long as at least one key in the quorum was uncracked. Using a fresh receive address every time is free, and in this case it was a real defense.
The lesson for anyone setting up a vault is plain. Multisig protected people whose keys came from different manufacturers. A three-of-a-kind vault moved the single point of failure from the seed to the factory.
Two things often mistaken for multisig
Two other tools also split up or harden a backup, and they are often recommended in the same breath. They solve different problems.
| Multisig (2-of-3) | Shamir backup (SLIP39) | BIP-39 passphrase | |
|---|---|---|---|
| What gets split | The signing power: three keys, any two sign | The backup of one key, into shares | Nothing; an extra secret word is added to one seed |
| Devices needed to spend | Two of three | One | One |
| Xpubs or descriptor to back up | Yes | No | No |
| Protects against a flawed device maker | Yes, if keys come from different makers | No, one device generates the key | Only if the passphrase is long and random |
| Single point of failure remaining | None by design | The one device, during setup and every time you spend | The one device |
| Can be changed without moving coins | No | Yes, per Trezor | No |
Shamir backup is not multisig. SLIP39 splits the backup of a single key into shares, so that, for example, any two of three shares rebuild it. Trezor recommends it as the simpler choice for individuals: one device, one signature, no xpubs, and it can be upgraded without moving funds. The trade-off, per Unchained, is that Shamir still creates temporary single points of failure during setup and whenever the key is reassembled to spend.
A passphrase is not multisig either. It adds a secret word to a single seed, so someone who finds the 24 words still cannot spend. Casa notes that at least one Mk3 wallet protected by a two-word passphrase was drained anyway. A short or guessable passphrase gives little protection.
One other point: a 2-of-2 or 3-of-3 setup is technically multisig, but ForkLog warns against both, because losing any single key locks the wallet for good. It adds failure points rather than removing them.
Building one yourself, or sharing it with a company
There are two ways to get a 2-of-3. You can hold all three keys and run free software as the coordinator, or you can pay a company to hold one key while you hold two. This second kind is called collaborative custody, and the company's key alone cannot move your coins.
Trezor Suite, Trezor's own app, does not support multisig, so Trezor owners pair their device with Electrum, Sparrow or Casa. Sparrow Wallet is free and open source. Nunchuk offers a free tier for fully do-it-yourself multisig with any quorum.
A DIY 2-of-3, in the order Coldcard's guide gives it
-
Initialize three signing devices
Each generates its own seed. Coldcard's guide recommends at least 50 physical dice rolls per cosigner and a unique, strong passphrase for each. Casa's figures agree: about 50 private rolls give roughly 128 bits of entropy, about 99 give roughly 256.
-
Export each device's xpub
Coldcard's guide uses a MicroSD card (FAT32 or FAT12, 512 MB or larger) and the BIP-48 native SegWit path, m/48'/0'/0'/2'.
-
Create the wallet on one device and register it on the others
Device 1 builds the P2WSH 2-of-3 wallet and saves an export file. Devices 2 and 3 import that same file. A device that has not registered the wallet should not sign for it.
-
Import the wallet into the coordinator as watch-only
Sparrow reads the export file and can then draft and broadcast transactions, but holds no private keys.
-
Compare the first receive address everywhere
Check that Sparrow and every signing device show the same first address before sending anything.
-
Send a small test amount, then spend it
Draft a PSBT in Sparrow, sign it with two devices, and broadcast. Only fund the wallet properly once a spend has worked.
-
Back up everything
Coldcard's list: three seed backups, each device's encrypted backup file, the wallet export file and a Sparrow wallet export. Do not change a cosigner's passphrase after registration; Coldcard warns it breaks the wallet's expected setup.
Where to buy the devices
A hardware wallet is only as trustworthy as the path it took to you. Buy each one from the maker's own store, or from the maker's own Amazon storefront where the listing says it is sold by the maker. Never buy from a third-party seller, never buy used, and never accept a device that arrives "pre-configured" or with a seed already written down. Generate every seed yourself, on the device. If you are still choosing a first device, the wider steps for moving coins off an exchange safely apply to each key in a multisig as well.
What each route costs
Prices below are as published on Unchained's comparison page and Spark.money's service comparison when this page was researched in November 2026. Neither page carries a date, so confirm on the provider's own site before signing up. Hardware is extra unless noted.
| Route | Setup | Annual cost | Key notes |
|---|---|---|---|
| Sparrow (DIY) | 2-of-3 or other, all keys yours | Free software | You hold every key and every backup; no one to call |
| Nunchuk free tier | Any quorum, DIY | $0 | Key replacement is manual |
| Nunchuk Iron Hand | 2-of-3, Nunchuk holds one key | $120/year | Lowest-priced paid option in the sources |
| Nunchuk Honey Badger | 2-of-4, you hold three keys | $480/year | |
| Nunchuk Honey Badger Premier | 3-of-5 | $2,100/year | Includes three wallets and two inheritance keys |
| Casa Standard | 2-of-3: phone key, hardware key, Casa recovery key | $250/year | Email only, no KYC; inheritance included with a 6-month wait; 7-day recovery wait |
| Casa Premium | 3-of-5: four client hardware keys, one Casa key | $2,100/year | Three hardware devices and Faraday bags included; 48-hour recovery wait; one free device replacement a year |
| Unchained standard vault | 2-of-3: two client hardware keys, one Unchained key | $250/year per vault | Full KYC; 1.00% trading fee, $2,000 minimum purchase |
| Unchained IRA vault | 2-of-3 inside an IRA | $250/year plus 1.50% per trade | |
| Unchained Signature | Up to 15-of-16 via Connections | $6,000 first year, $4,500 renewal (personal); $7,500/year business | Two premium hardware wallets, relationship manager, inheritance with no waiting period; 0.75% trading fee |
For the DIY route, the main cost is hardware. Anant Tapadia, writing in a Bitcoin Magazine guest column, put a multi-vendor 3-of-5 at roughly $250 to $600. A 2-of-3 needs three devices instead of five. This page has no current per-device prices from the makers' own sites, so check those directly.
The brief for this page also asked about Xverse with Asigna and BitPay's wallet. None of the sources carries their current multisig pricing or setup details, so they are left out rather than guessed at.
The coordinator is the part that can lie to you
Your signing devices keep the keys offline, but the coordinator on your computer is what shows you addresses and drafts payments. If it is malicious or compromised, it can try to trick you. Tapadia lists five ways this can happen: showing a wrong receive address (one the attacker controls), swapping the destination of a payment, replacing the change address so the leftover from a payment goes to the attacker, altering the wallet registration so the devices sign for a wallet you did not set up, and a ransom attack.
The ransom case is the subtle one. An attacker who controls fewer keys than the threshold cannot spend. But if you later lose enough of your own keys that you can no longer reach the threshold without theirs, they can demand payment for their signature. A coordinator that sneaks its own key into your setup is how this begins.
-
Verify every receive address on a signing device's screen, not just the computer
The device knows the registered wallet and cannot be fooled by the computer's display.
-
Register the wallet on all signing devices and confirm the details on each
Check the threshold (2 of 3) and that the key fingerprints match the three devices you own.
-
Check the destination and change address on the device before signing
A device that has registered the wallet can confirm change is going back to you.
-
Cross-check the wallet setup in a second coordinator
Tapadia suggests confirming the configuration in more than one piece of software.
-
Prefer open-source, standards-based software
Support for PSBT and BSMS (Bitcoin Secure Multisig Setup) makes setups portable and verifiable.
-
Test with a small amount before funding
Receive, then spend, before the real deposit.
-
Never reuse an address
The 2026 Coldcard thefts showed reused addresses were the first to go.
So, should you set one up?
No source gives a dollar threshold, and this page will not invent one. What the sources do offer is a set of questions, and a rough way to think about spending.
Theya's help center asks the question that matters most: "If you're holding bitcoin for 4-10+ years, are you satisfied with your current custody setup?" It says the more Bitcoin matters to your long-term finances, the more security you should want, and that a common path is to start with single-sig and upgrade as holdings grow. Tapadia's budget test is that spending under 10% of holdings on security is reasonable for about 0.5 BTC, which he valued at roughly $11,000 when he wrote it. Unchained and Tapadia both say multisig suits people who withdraw rarely; it is a vault, not a checking account.
| Your situation | Single-sig + good backup | Single-sig + SLIP39 | DIY multisig (Sparrow, Nunchuk free) | Collaborative (Casa, Unchained, Nunchuk paid) |
|---|---|---|---|---|
| Small share of net worth, one owner, still buying | Best fit | Fine if your device supports it | More work than the amount justifies | Fees are large relative to holdings |
| Large enough to change your retirement, one owner, 4 to 10+ years | Single point of failure | Better backup, same single device | Good fit if you will verify every step | Good fit if you want help recovering a lost key |
| Shared with a spouse or business partner | Hard to share safely | Shares can be split, but one device signs | Good fit: each person holds a key | Good fit |
| Planning for heirs who are not technical | One seed must be explained and handed over | Shares can be spread among family | Heirs must rebuild a wallet from five backups | Strongest fit: the company can help an executor |
| Bitcoin held inside an IRA | Not available in an IRA | Not available in an IRA | Not available in an IRA | Unchained offers IRA vaults |
Inheritance deserves its own line. Jeff Vandrew of Unchained makes a point that applies to every setup: controlling the keys is not the same as owning the coins. "Possession alone, however, does not create valid title," he writes. Bitcoin passes under a will like any other property, and an executor still has to inventory it. Where multisig helps is the handoff. In a 2-of-3, "someone with access to just one key cannot spend your bitcoin alone," so a key can sit with an executor or trustee while you are alive without any risk. Where it hurts is complexity: an heir has to understand devices, seeds and a descriptor. The page on planning for a hardware wallet covers the will and letter of instruction side.
Whatever you choose, the seeds are still words on something that can burn. Each of the three belongs on a steel backup in its own place, with a copy of the descriptor beside each one.
Not yet holding your own keys at all?
The full self-custody walkthrough covers moving Bitcoin off an exchange, ETF or IRA and storing the first seed properly.
Read the self-custody guide
Comments
No comments yet. Be the first to comment!
Leave a Comment