Skip to content

Safety

Short answer: Stray cannot take your funds. A scan only reads. A claim is a transaction your own wallet sends, limited by checks that run on-chain and in your browser. Here is each of them, in enough detail to check for yourself.

The scan only reads

A scan needs your public address and nothing else. It reads public chain data. There is no wallet connection, no signature and no permission involved, which is why you can scan any address, including ones you do not own.

The claim runs from your wallet

Claims are transactions your own wallet sends. We prepare the list of calls; your wallet shows them, and nothing happens unless you sign. Every call is labeled in plain language, for example "Collect fees from Uniswap V3 #48211" or "Service fee 0.0171 ETH".

Two ways to claim, chosen per chain

One atomic batch. If your wallet reports atomic batching (EIP-5792) on a chain, the calls for that chain go into a single batch. If any step fails or returns less than its minimum, everything reverts. An atomic batch is never split up and sent call by call.

Step by step through the Stray router. Wallets without atomic batching (Rabby and most plain accounts) send ordinary transactions one after another: the claims, exact approvals to the Stray router, and one settlement transaction. Here the steps are separate: a claim can be on-chain while the settlement is still to come. That is the honest trade-off, and the rest of this page says what still protects you. If neither way works on a chain, it is not offered with that wallet.

BalanceGuard: the on-chain check (atomic batch)

Every atomic batch is wrapped by a small contract called BalanceGuard:

  1. Before the claims, it records your wallet's balance of what you are about to receive.
  2. After the claims and any sales, it compares. If your balance did not rise by at least the amount the simulation proved, it reverts, and the whole batch reverts with it.
  3. Only then does the fee transfer run.

This means the fee can only ever be paid in a transaction in which the money really arrived, in that same transaction. The guard holds no funds; it only reads balances and reverts.

The Stray router: the on-chain check (step by step)

In a step-by-step claim the last transaction calls the Stray router, a contract without owner or admin keys that holds no funds between transactions:

  1. It pulls exactly the approved amount of each token you sell into an executor contract that belongs to your wallet alone, which sells it on an allowlisted venue with a fixed minimum.
  2. It measures what the sales really returned. If that is less than the minimum set from the simulation, the transaction reverts and nothing is sold.
  3. Only then does it split the fee from what it measured, at most 5%, and send the rest to you (or to the address you chose).

Before the settlement is sent, your browser also checks that the claimed tokens really arrived, so a claim that paid nothing can never lead to selling tokens you already held.

Only known contracts can be called

A claim plan may call two kinds of contract and no others:

  • Claim targets: the protocol contracts registered for each kind of finding, such as a position manager or a fee locker.
  • Swap routers on a fixed allowlist, used only when you choose to convert tokens, each swap with a minimum output.
  • The Stray router (step-by-step claims only), at one fixed address. Your browser decodes its call completely: fee address, fee rate, referrer, recipient, deadline, and every sale it will make.

Your browser checks every plan against the same lists before your wallet is asked to sign. A plan that calls anything else, pays the fee to any address but ours, or charges anything other than 5% is refused in the browser, before it reaches your wallet.

Approvals are exact and short-lived

In an atomic batch the approval goes to the swap router for that sale and is used up in the same transaction. In a step-by-step claim the approval goes to the Stray router, for the exact amount the settlement will pull, never unlimited, and the settlement uses it up. An allowance that already covers the amount is reused instead of asking again. Stray never asks for unlimited approvals, and no approval names an address of ours that could move tokens on its own.

Re-simulated right before you sign

The full batch, or the whole step-by-step sequence in order, is simulated again at the latest block before it is sent to your wallet. If amounts changed, the plan changes with them. If a finding stopped being claimable, it drops out.

On Solana

Solana has no batches that revert as a whole, so a Solana claim is built from independent transactions, each safe on its own:

  • Every transaction carries its own value and its own fee. A fee is never sent in a separate transaction from the value it is charged on.
  • Only known programs. Your browser decodes every instruction before your wallet is asked to sign. Programs must be on a fixed list (the token programs, Jupiter, and the protocols the claims belong to). Token instructions may only close accounts back to your wallet, burn tokens you chose to clear out, or pay the exact fee to our fee address. Approvals, delegates, authority changes and transfers to anyone else are refused.
  • Simulated twice. We simulate every transaction right before you sign and check the balances: your wallet must receive at least the amount shown, our fee address exactly the fee, and none of your token accounts may lose anything beyond what you chose to sell or burn. We check again on-chain after it lands.
  • Checked by our server before sending. The signed transaction must be exactly the one we prepared. A transaction that already landed is never sent again, and signing a step twice can never pay twice.
  • Gasless claims add our relayer as the fee payer. It signs only the exact transaction you reviewed, after a fresh simulation, and can only pay the network fee and deposits for your own token accounts. It can never move your funds, and you sign nothing for it.

What it means in practice: if you stop halfway, the transactions that landed are complete claims with their fee; the ones that did not land cost nothing.

Your browser runs its checks and simulations through the Solana RPC configured for this site (NEXT_PUBLIC_SOLANA_RPC_URL). That RPC is independent third-party infrastructure, not operated by Stray, so a compromised Stray server cannot also fake what your browser sees.

What we will never do

Ask for a seed phrase or private key
Not in the app, not in support, not by email. Anyone who does is not us.
Ask you to sign a permission
No Permit, Permit2 or setApprovalForAll signatures. The only message you sign is a free sign-in.
Contact you first
We do not send DMs offering recovery. Offers that arrive unasked are scams.
Hold your funds
Claims go from the protocol to your wallet. Nothing passes through an account of ours.

If your wallet is compromised

If someone else controls your keys, they can claim faster than you. Do not use Stray from a compromised wallet; move what you can to a fresh wallet first and contact support.

Checking it yourself

  • Open any call in your wallet before you sign: the target is the protocol's contract, the guard or the Stray router, never an unknown address.
  • Read the batch on a block explorer afterwards: the guard's check sits between the claims and the fee. For a step-by-step claim, the settlement transaction's Swept event lists what was sold, the fee and what you received.
  • Scan first, claim later. Nothing about scanning commits you to anything.