What is intent?
A signed statement of the outcome a user wants, such as a specific amount of one asset for another, left to competing parties to fulfill however they can.
Not yet verifiedHow we verify
3 min read
In this entry
A signed statement of the outcome a user wants, such as a specific amount of one asset for another, left to competing parties to fulfill however they can.
It inverts the usual model, where a user specifies the exact route and transaction. Intent systems can find better prices, batch orders together, and shield users from some value extraction, because the fulfilling party absorbs the execution risk. The tradeoff is dependence on the network of fulfillers.
You meet the model without the vocabulary: a swap interface that quotes you a guaranteed output and charges nothing if it fails, or a cross-chain bridge that delivers on the destination chain in seconds. Both are usually intents underneath.
How it works
In the conventional model you build a transaction that names the pool, the route, and the minimum output, then sign and broadcast it. Everything between signing and inclusion is your problem: the price can move, the route can fail, and a searcher can sandwich you.
An intent moves that burden. You sign a message stating the constraint you care about, for example at least 1,950 USDC for 1 ETH, valid for two minutes. That message goes to a network of solvers rather than straight to the chain. Solvers compete to satisfy it, using their own inventory, several venues at once, or a batch that nets your order against someone trading the opposite way. The best result wins, and the solver submits and pays for the transaction.
The consequences are practical. Your output is either at least what you specified or the intent expires unfilled, so a failed attempt does not usually cost gas. Batching lets orders cross directly without touching a pool at all. And because you are not broadcasting a route, the usual sandwich attack has nothing to target.
What you take on instead is a new dependency. Fulfillment relies on solvers choosing to act, so a small or illiquid intent may simply expire. Solvers keep the difference between what they achieve and what they promised you, which is a spread you do not see itemized. And the settlement contract holding the process together is code like any other, with the risks that implies.
Example
Illustrative arithmetic. You sign an intent to sell 1 ETH for at least 1,950 USDC. Three solvers compete. One routes through a single pool and can achieve 1,958. One splits across two venues and reaches 1,962. A third finds another user simultaneously buying 1 ETH and matches you directly, avoiding pool fees entirely, reaching 1,971. The third wins and you receive 1,971 USDC, or possibly 1,965 with the solver retaining the rest. Under the conventional model you would have picked one route yourself and taken whatever it gave, minus whatever a sandwich extracted.
Why it matters when you buy
This is a decentralized exchange pattern, and it changes what a quoted price means: an intent quote is a floor you are promised rather than an estimate that can slip. On centralized venues the comparable question is the spread and depth of the actual book, which the liquidity view measures per exchange, and the fee comparison covers all-in cost. Centralized versus decentralized exchanges covers the wider tradeoff.
Related terms
- solver: the competing party that fulfills an intent
- mev: the extraction this model reduces
- sandwich attack: the specific attack it defends against
- slippage tolerance: the manual equivalent constraint
- account abstraction: related flexibility in how transactions are authorized
- dex: the venues solvers usually route through
Questions
Do I pay gas if the intent is not filled?
Generally no. The solver submits and pays for the transaction, so an unfilled intent typically costs nothing beyond the time it was open.
How do solvers make money?
By keeping the difference between the outcome they achieve and the minimum they committed to you, plus any protocol rebate. That margin is real and is not itemized in the quote.
Is this safer than a normal swap?
It removes some execution risks and adds a dependency on solvers and on the settlement contract. Different risk, not uniformly less of it.