A trader has moved several thousand dollars worth of Ethereum and stablecoins into a new wallet application. The interface is clean, the gas fee estimates are clear, and integration with decentralized exchanges feels seamless. But a legitimate question arises: if the wallet provider could theoretically access the recovery phrase or intercept a transaction, is the wallet actually non-custodial? The answer depends on understanding what non-custodial means, how keys are stored, and which parts of the system the user controls versus which parts depend on trust.
The distinction matters more than marketing language suggests. A wallet can call itself self-custodial while still introducing points where private keys could be exposed, modified, or stolen. Conversely, a genuinely non-custodial wallet still requires the user to secure a recovery phrase, understand the application’s code, and verify network connections. Rabby Wallet has positioned itself as a self-custodial option for Ethereum and EVM-compatible networks, but the practical meaning of that claim deserves careful examination rather than acceptance at face value.
How non-custodial actually works at the technical level
Non-custodial means the wallet provider does not hold the user’s private keys on company servers. Instead, keys are generated, stored, and used locally on the user’s device. When a transaction is signed, the cryptographic operation happens within the wallet application, and the signed result is then broadcast to the blockchain. The provider never sees the unencrypted key material. This is fundamentally different from a custodial exchange, where users send funds to an address controlled by the exchange and trust the company to hold and return the balance.
Rabby’s architecture follows the non-custodial model. When a user creates or imports a wallet, the seed phrase and derived private keys remain stored locally on the device, encrypted with the user’s chosen password. The browser extension or mobile app performs key derivation and transaction signing internally. No unencrypted keys are transmitted to Rabby’s servers or any third party. From a cryptographic standpoint, Rabby cannot unilaterally move user funds because it never obtains the signing material required to do so.
However, non-custodial is not the same as trustless. The user still trusts that the wallet software is not malicious, that the device itself has not been compromised, and that the recovery phrase remains secret. A compromised browser extension, for example, could theoretically extract keys from memory or intercept them during the signing process. A weak password could allow offline brute-force attacks against the encrypted key store. The non-custodial property means the risk is not centralized with Rabby as an institution; instead, it is distributed across the user’s device security and the integrity of the wallet software itself.
The open-source nature of Rabby’s browser extension code is relevant here. Open source allows independent security researchers to audit the code and identify backdoors, logging routines, or key-stealing logic that a closed-source wallet might hide. That transparency is valuable for reducing the attack surface, but it does not guarantee security. An audited version of the code could still be modified during compilation or distribution, or a user could install a counterfeit version from an untrusted source. The defense is to verify code, check installation sources, and understand that open source is a necessary condition for trust but not sufficient on its own.
Private key storage and encryption: where control actually lives
The moment a user creates or imports a wallet into Rabby, they encounter the backup recovery phrase—typically 12 or 24 words that can regenerate all private keys associated with the account. This phrase is the critical security object. If someone obtains it, they can recreate the wallet on any device and spend all associated funds. Rabby cannot and should not store this phrase on its servers. Instead, the user is responsible for writing it down, photographing it, encrypting it, or storing it in a physical vault. That responsibility is the core of self-custody.
Within the application, Rabby encrypts the recovery phrase using the user’s password. That encryption is local; the encrypted data stays on the user’s device and in whatever cloud backup service the device uses (for mobile apps), if the user enables it. The encryption key is derived from the password, which only the user knows. A strong password is essential because the security of the encrypted key material depends on it. An attacker who gains access to the encrypted data but not the password cannot easily extract the key material. Conversely, a user who loses the password and the backup recovery phrase has irreversibly lost access to the funds.
The tradeoff between convenience and security is therefore unavoidable. Rabby offers account recovery options that might include confirming recovery phrases or importing from hardware wallets such as Ledger or Trezor. When using a hardware wallet, the signing operation happens on the hardware device itself, and the private key never enters the computer at all. That is the strongest configuration for security but introduces friction for frequent transactions. For users keeping funds in a Rabby wallet directly, the strength of the device password and the secrecy of the recovery phrase are the decisive factors.
Users should also be aware that if the device is lost or compromised, the risk profile changes. Malware on a computer could observe key material during signing or extract the encrypted key store. A phone that has been jailbroken or rooted may have reduced isolation between applications. A device stolen before the password is entered is safer than one stolen with a logged-in session. These are device-level security considerations that Rabby can influence through best practices but cannot entirely eliminate.
Transaction signing and network interaction
When a user creates a transaction in Rabby—such as sending tokens or interacting with a smart contract—the signing process is local. The wallet constructs the transaction data, the user reviews it, the private key signs it without leaving the device, and the signed transaction is broadcast to a blockchain node. Rabby may use public blockchain infrastructure such as Infura, Alchemy, or QuickNode to submit transactions, but these services receive only the signed data. They cannot reverse-engineer the private key from a signature, nor can they modify the transaction once it is signed without making it invalid.
This is where transaction simulation becomes important for practical security. Rabby displays what a transaction will do before signing—showing token transfers, contract function calls, and potential risks. A user can detect that a transaction is attempting to approve an unintended contract or send funds to the wrong address. That human review step is a critical control point, but it depends on the transaction preview being accurate. If the wallet displays misleading information, the user could approve something harmful.
The network nodes that receive the signed transaction do not know which user sent it if proper privacy practices are followed. However, the sender’s IP address is visible to the node unless the user routes through Tor or a VPN. Rabby does not provide built-in privacy networking; that is a separate concern from non-custodial key management. A user concerned about link-ability between transactions and identity should use additional tools for network privacy, not expect the wallet alone to provide it.
Another practical issue is the recovery of stuck transactions. If a transaction is submitted with insufficient gas or if network congestion delays confirmation, the user can potentially replace it with a higher-fee version. Rabby displays transaction status and can facilitate replacement, but the user must retain control of the signing step. If the wallet could independently resubmit or accelerate transactions without user approval, that would be a custody claim—signing on behalf of the user. Rabby does not do this, preserving the user’s control.
Hardware wallet integration and custody boundaries
One of Rabby’s significant features is compatibility with hardware wallets. Ledger and Trezor devices can be used to sign transactions while keeping private keys on the hardware device entirely. In this configuration, Rabby functions as an interface to submit transactions to the hardware wallet, and the hardware device approves or rejects them. The user reviews the transaction on both the Rabby screen and the hardware device’s display, creating a secondary verification point.
This setup is the strongest custody model available because the private key never enters the computer. Malware on the computer, a compromised Rabby installation, or intercepted network traffic cannot sign transactions without physical approval from the hardware wallet. The user controls access through a PIN on the device, and the recovery phrase is stored only on the hardware wallet itself. If the computer is entirely compromised, the funds remain safe as long as the hardware device and PIN are secure.
The tradeoff is friction. Signing every transaction requires physically interacting with the hardware device, confirming details on its small screen, and waiting for the confirmation. For users moving large sums or making infrequent transactions, this is acceptable. For traders executing rapid swaps or users managing multiple daily transactions, the overhead can become impractical. Many users choose hardware wallets for larger holdings and keep smaller amounts in Rabby for active trading.
Rabby’s support for hardware wallet import is therefore not a limitation of non-custody but rather an optional upgrade in custody strength. A user can start with a Rabby-native account, later migrate to hardware wallet control, or maintain both simultaneously. The flexibility means custody boundaries can be adjusted as the user’s risk tolerance and activity level change.
The MetaMask import question and migration risks
Rabby allows users to import accounts from MetaMask, which is straightforward when both wallets are used by the same person. However, an imported account in Rabby is still the same account—if the MetaMask seed phrase is compromised, both wallets are compromised. The import does not create a new custody relationship; it simply allows the user to access the same private keys through a different interface. For this reason, account import should only be used when the original wallet is fully secured or when the user intends to migrate fully and deprecate the old wallet.
If a user imports a MetaMask account into Rabby and continues to use MetaMask as well, they are maintaining two interfaces to the same key material. Keeping multiple applications in sync with the same keys introduces more attack surface. The safer practice is to import only when migrating deliberately and to retire the old wallet after confirming the new one is working correctly. This is a user discipline issue rather than a deficiency in Rabby’s non-custodial design, but it is a risk that the import feature creates.
When importing, users should verify that the imported account displays the same address and balances in Rabby as in MetaMask. Mismatches could indicate incorrect derivation paths or a wrong seed phrase. After importing, the recovery phrase for the account should be known and secured. If the original MetaMask seed phrase is unknown, the user cannot securely retire MetaMask and relies entirely on Rabby, which means all security depends on Rabby’s password encryption and device security.
Multi-chain complexity and where custody boundaries blur
Rabby supports Ethereum, Base, Arbitrum, Optimism, Polygon, BNB Smart Chain, Avalanche, and other EVM-compatible networks. All of these use the same underlying key derivation—a single seed phrase generates private keys that work across all EVM chains simultaneously. This is efficient and reduces the number of recovery phrases a user needs to manage, but it also means that if a recovery phrase is compromised, an attacker gains access to funds across all supported networks.
The automatic network detection feature in Rabby is convenient—when a decentralized application switches networks, Rabby adjusts accordingly. However, this automation also creates a risk: a user could accidentally approve a transaction on the wrong network if they are not paying close attention. Swapping tokens on one chain is very different from the same action on another, as fees, available liquidity, and counterparties differ. A transaction review habit must include checking which network is active before signing.
Bridging tokens between chains introduces additional complexity. When moving funds from Ethereum to Arbitrum, for example, the transaction involves a bridge contract that locks tokens on one chain and mints an equivalent amount on another. The bridge provider controls the locked tokens temporarily, introducing a custodial element. Rabby displays the bridge operation in the transaction simulation, but the user is trusting the bridge protocol itself. This is not a limitation of Rabby’s non-custodial design—it is inherent to cross-chain operations. Understanding that custody risk shifts from Rabby to the bridge provider is essential.
Smart contract interaction risks and the limits of self-custody
Rabby’s transaction simulation is powerful because it attempts to decode what a smart contract call will actually do. Rather than showing raw bytecode, the wallet displays something like “Approve 100 USDC to contract 0x1234…” or “Provide liquidity and receive LP tokens.” This is a significant usability improvement over raw transaction data, and it helps users avoid obvious mistakes such as approving their entire balance to an unfamiliar contract.
However, the simulation is only as good as the contract interpretation available to Rabby. A contract that hides its true function, uses unusual encoding, or has been recently deployed might not be decoded correctly. A user who does not understand what “approve unlimited spending” means—that the contract can take up to the approved amount whenever it wants—could end up losing funds to a malicious contract later. Rabby’s transaction preview reduces but does not eliminate this risk. Self-custody includes the responsibility to learn what operations mean, not just to sign whatever the interface displays.
Contract risks are also time-delayed. A user might approve a contract that is trustworthy today but becomes compromised later through an exploit or upgrade. Revoking approvals periodically or using tools that show active approvals can help manage this. Rabby provides visibility into token approvals, which is useful, but the user must actively manage them. Non-custodial custody means taking responsibility for transaction decisions, not merely holding keys.
What self-custody demands from users in practice
Non-custodial operation places a higher burden on users than custodial services. A centralized exchange will reset a forgotten password through identity verification, can potentially recover stolen funds if the user reports fraud quickly, and maintains insurance policies. Rabby can do none of these things. A lost password combined with a lost recovery phrase means permanent loss of access. A stolen recovery phrase means permanent loss of funds, with no recourse. These are the consequences of actual self-custody, and they require discipline.
Practical self-custody means creating the recovery phrase in a secure environment, storing it offline in multiple secure locations, using a strong device password, enabling any available device-level security features such as biometric authentication or TPM-backed encryption, and treating the browser extension or mobile app as a tool that should be kept updated. It means reviewing transactions before signing, understanding the difference between mainnet and testnet, and being cautious when connecting to decentralized applications or entering recovery phrases anywhere.
The security decision is not simply whether to use Rabby but whether to invest in the practices required to use it safely. A user who writes the recovery phrase on a sticky note and uses password “123456” cannot claim to be practicing self-custody effectively, regardless of how non-custodial the wallet architecture is. Conversely, a user who secures the recovery phrase, maintains device hygiene, and uses a hardware wallet for large holdings has made a genuine commitment to non-custodial control.
Rabby’s role in that system is to not interfere. By keeping private keys local, avoiding compulsory cloud backup, providing clear transaction previews, and supporting hardware wallets, the wallet reduces the friction between genuine self-custody and convenient operation. But the wallet cannot substitute for user responsibility. Non-custodial does not mean worry-free; it means the worry is distributed correctly—between the user’s own actions and the technical implementation, not between the user and a company that could theoretically move their funds.
Frequently asked questions
Can Rabby access my private keys or funds?
No. Rabby is non-custodial, meaning your private keys are generated, stored, and remain on your device. Rabby servers never see unencrypted key material. The wallet cannot move your funds without your signature, and signing happens locally within the application. Your security depends on your device password, recovery phrase secrecy, and the integrity of the wallet software itself, not on trust in Rabby as an institution.
What happens if I lose my recovery phrase?
If you lose both your device password and your recovery phrase, you have permanently lost access to the funds in that wallet. There is no password reset, no account recovery, and no customer support intervention. This is the direct consequence of self-custody. The recovery phrase is the only way to restore access on a new device. Store it offline in multiple secure locations before any crisis occurs.
Is using Rabby with a hardware wallet safer than using Rabby directly?
Yes. A hardware wallet keeps your private key on a separate device and signs transactions there, so the private key never enters your computer. Even if Rabby or your computer is completely compromised, the hardware wallet can only be controlled with physical approval and a PIN. For large holdings, hardware wallet integration is the strongest custody configuration available.