What are intentions and solvers? The invisible layer for executing DeFi transactions

Introduction

Most DeFi users believe that when they redeem tokens on Uniswap or SushiSwap, they interact directly with an automated market maker (AMM). In 2022, this situation is basically true. A user signs a transaction, which goes into a public memory pool, the validator packs it into a block, and AMM's constant product formula determines the price.

There is a problem with this model. Public memory pools are hunting grounds. MEV robots monitor pending transactions and perform pincer attacks: they buy before your trade drives up the price, and then sell later, extracting value from the spread. Flashbots estimates that by 2023, cumulative withdrawal profits from MEVs on Ethereum will exceed US$600 million, a large portion of which will come from a pincer attack on retail exchanges.

Intent-based protocols refactor this process. Instead of broadcasting transactions that specify all execution details, users sign an intent: a declarative statement of the expected outcome. A network of solvers then competes to fill in that intent at the optimal price, off-chain, and without exposing the order to a common memory pool.

This guide will explain how intents and solvers work, how the main protocols implement them, and the tradeoffs that users need to accept.

Question of directly interacting with AMM

When users redeem tokens through traditional AMM, the transaction codes a specific path: exchange token A for token B on pool X, with a minimum output of Y and an expiration time of Z. This specificity raises three problems.

MEV vulnerability

Transactions wait in the public memory pool until the verifier packages them. During this period, robots can see and attack expected transactions, extracting value from users. Academic research records from the Flashbots team show that on Ethereum alone, flanking attacks cause approximately $200 million to $300 million in losses to retail users every year. In one particularly high-profile case, a DeFi trader suffered a 100% slip in a pincer attack and lost the full value of the trade.

Routing is not optimal

Users submitting transactions to a single AMM can only receive the price of that AMM. But liquidity is scattered across dozens of DEXs, multiple chains and centralized exchanges. The best price for a given redemption may involve splitting the order into three pools on two chains, a path that users have never considered for simple transactions.

Gas inefficiency

Each user pays Gas fees separately. If 50 users wanted to convert ETH to USDC in the same block, they would submit 50 separate transactions, each paying for their own Gas expenses. There is no batch processing mechanism.

How intention works

Intention reverses the trading model. Instead of specifying how to execute transactions, users specify what they want to achieve.

A typical intention includes:

Enter token and amount: the asset the user is willing to spend.
Export tokens and minimum quantity: The assets the user wants to receive, with a guaranteed price.
Expiration time: The point in time at which the intention expires.
Signature: Cryptographic proof that the user authorized this intention.

The intention is not a blockchain transaction. It is an off-chain signed message that is submitted to the protocol-specific order flow system rather than the Ethereum memory pool. This difference is crucial: because intentions never enter the public memory pool, MEV robots scanning for pinch-in opportunities are invisible.

Once submitted, the intention is to enter a solver auction. The signing message grants conditional approval to the settlement contract, allowing it to transfer the user's input tokens, but only if the solver delivers the promised output. Users 'funds remain in their wallets until atomic settlement is made.

What a solver does and how it competes

A solver is an entity (robot, market maker, trading company) that monitors incoming intents and competes to fill them.

The competition mechanisms vary among different protocols, but the overall structure is:

Intention Broadcast: The protocol distributes new intents to registered solvers.
Scenario generation: Each solver analyzes the intent and decides how to fill it. The solver may route through multiple DEX pools, leverage private inventory, bridge from another chain, or merge multiple intentions into one batch.
Bid Submission: The solvers submit their proposed execution plan, including the output the user will receive.
Auction ruling: The agreement selection winning solver is usually the one that provides the user with the best price after all costs.
On-chain settlement: The winning solver executes transactions on the chain, and the settlement contract verification user receives at least the promised minimum output.

The solver assumes its own risk. They advance funds, pay Gas and deal with execution complexities. Their profits come from the spread between the price they can get and the price they bid to users, minus Gas and capital costs. Competition among solvers compresses this profit margin and returns more value to users.

The economics of solver operations create natural barriers to entry. Competitive solutions require money for inventory, low-latency infrastructure to monitor multiple sources of liquidity, and complex routing algorithms. Solvers do the same job, but direct value to the user rather than extract it from the user.

CoW Protocol: Batch auctions and demand coincidence

CoW Protocol (formerly CowSwap) pioneered the intention-solver model on Ethereum. Its name comes from the economic concept of "Coincidence of Wants."

The key innovation is bulk auctions. Rather than filling orders one at a time, the CoW Protocol collects intentions over a window of time (about 30 seconds) and then runs a single batch auction in which the solver competes to fill all orders simultaneously.

This creates a direct match opportunity. If Alice wants to sell 1 ETH in exchange for USDC and Bob wants to buy 1 ETH with USDC, then the solver can match them point-to-point without touching the liquidity pool. Neither party is required to pay AMM fees or price differences. The solver profits by capturing the spread between two users 'limit orders.

CoW Protocol calls this transaction a "coincidence of demand" transaction. In practice, pure CoW transactions account for a meaningful minority in transaction volume, but when they occur, both parties get a better price than any AMM.

For orders that cannot be matched point-to-point, the solver routes them through on-chain liquidity. The batch auction format is still helpful: Because all orders are settled in a single transaction, Gas costs are spread across the batch. A batch of 30 conversions pays only one fixed transaction expense instead of 30.

By mid-2026, CoW Protocol has processed more than US$80 billion in cumulative transaction volume, making it one of the largest DEX protocols by number of transactions. Its solver team has also matured, with mature market makers and trading companies competing with independent solver operators.

UniswapX: Uniswap's intent layer

UniswapX will be launched in 2023, adding an intent-based execution layer to Uniswap's existing liquidity pool.

When users submit a redemption through the Uniswap interface, they can choose to use UniswapX. The redemption is no longer routed directly through the Uniswap V3 or V4 pool, but becomes an intention. Solvers (called "fillers" in UniswapX terminology) compete to fill it.

UniswapX introduces Dutch-style order auctions. The lowest output acceptable to users starts at a high price and decays according to a predetermined curve over time. The first solver willing to fill at the current price wins. This mechanism encourages the solver to fill up quickly (for better profits early on), while protecting users from getting bad prices (auctions start at aggressive levels).

A key design choice: If the solver is not filled before the order price drops to the Uniswap pool price, the order will automatically fall back to the on-chain Uniswap route. Users can always get at least the AMM price. The solver can win only by offering a better price.

UniswapX also introduces cross-chain intentions. Users on Arbitrum can express an intention to receive tokens on Optimism. The solver handles bridging, verifying the results of settlement contracts on both chains. From the user's perspective, this is a single redemption.

Across: Intention for Cross-Chain Transfers

Across Protocol applies the intent-solver model specifically to cross-chain Transfers.

Traditionally, bridging tokens between chains involves lock-in and casting mechanisms, optimistic verification windows (typically 7 days for optimistic Rollups), or bridges based on liquidity pools. All of these are slow, expensive, or both.

Across refactors the bridge into an intention. The user signs a message: "I have 1,000 USDC on Ethereum and want 1,000 USDC on Arbitrum." A solver (called a "repeater" in Across) immediately sends 1,000 USDC from its own inventory to users on Arbitrum, and then later requests reimbursement from Across's settlement system on Ethereum.

Result: The bridging time is shortened from minutes or days to seconds. Users do not need to wait for cross-chain verification. The solver assumes this waiting risk in exchange for a fee.

Across's verification layer uses the Optimistic Oracle (UMA). If no objection to the solver's claim is raised within the challenge window, the reimbursement is processed. This creates an economic game in which honest relays are profitable and fraudulent claims are punished by forfeiting deposits.

ERC-7683 and standardization advancement

A major limitation of the current intent system is that each protocol defines its own intent format, solver network, and settlement contracts. Intention submitted to CoW Protocol cannot be populated by the UniswapX solver. This fragmentation limits solver competition and reduces the pool of liquidity available for each system.

ERC-7683 was proposed by Uniswap and Across in 2024 to create a common standard for cross-chain intentions. The proposal defines a common intent format (called "CrossChain Order") that any protocol can adopt. Solvers that implement this standard can fill in intentions from any compatible protocol, increasing competition and improving prices.

The standard defines two interfaces: ISetlementContract (settlement contract implementation) and IOriginSettler/IDestinationSettler (handling cross-chain execution). By standardizing these interfaces, ERC-7683 will allow a single solver to operate simultaneously on CoW Protocol, UniswapX, and Across, competing for transaction flows from all three protocols.

Adoption is still in its early stages. The standard requires existing agreements to modify their settlement contracts, which involves security audits and governance voting. But the direction is clear: Intent-based transactions are moving towards a unified market for solvers rather than fragmented protocol-specific pools.

Tradeoffs in intent-based systems

Intent-based protocols improve user experience in terms of price and MEV protection. They also introduce new trust assumptions and risks.

Solver centralization

In practice, a few well-funded solvers win most auctions. CoW Protocol's solver rankings show that three to five solvers handle most of the transaction volume. If competition for the solver weakens, users will lose the price improvements that make the system valuable.

Delay

Batch auctions and solver competition add time between order submission and execution. CoW Protocol batches are settled approximately every 30 seconds. UniswapX's Dutch-style auction was resolved faster, but still involved delays. For time-sensitive transactions, this delay can be important.

Trust in the solver

Users trust the settlement contract to correctly perform the minimum output. Smart contracts are audited, but they are still smart contracts. In addition, off-chain auction mechanisms must be fair. If the protocol operator can privilege certain solvers, then the auction is not really competitive.

Review Risk

Since intentions are submitted offline to protocol specific systems, agreement operators can theoretically review certain intentions. This is a different trust model than submitting transactions directly to the Ethereum censory-resistant memory pool.

Regulatory concerns

The solver network that routes transaction flows is structurally similar to broker-dealers in traditional finance. The question of whether solver activities constitute regulated market makers remains unresolved. Regulatory clarity either legalizes the model or imposes compliance requirements, thereby reducing the number of active solvers.

What is not covered in this article

This guide explains the core mechanisms of the intent and solver. It does not cover:

Technical implementation of specific solver algorithms
Regulatory considerations surrounding solver activities (Order flow payment, Best enforcement obligations)
CoW Protocol or UMA Detailed token economics by (Across)
Relationship between Intention and the Ethereum Proponder-Builder Separation (PBS) roadmap

Practical check before using intent-based protocols

Compare prices

Before submitting the intent, compare the quote output to the direct AMM execution. Intent-based agreements should continue to provide better prices. If not, the solver auction may not be competitive.

Understand the fallback mechanism

If there is no solver to fill the order, UniswapX will fall back to the AMM route on the chain. CoW Protocol will expire unfilled orders. Understand what happens if the solver does not implement your intentions.

Check slip point tolerance

The minimum output function in the intention is similar to slip point tolerance. Setting it too tight may prevent the order from being filled. Setting too loose will give the solver room for worse prices. Most interfaces set default values, but users can adjust them.

Verify settlement contracts

Enforcement of intended smart contracts is a key trust component. Check whether it has been audited and by whom. Look for contracts that are immutable or managed by time-locked multiple signatures, rather than a single administrator key.

Consider order size

is intended to provide the greatest price improvement for medium to large orders because route optimization and MEV protection are the most effective. For very small conversions on the low-Gas chain, the price improvement may be minimal because MEV withdrawals are less profitable on small orders.

Note Gas overhead

On the Ethereum main network, the execution of settlement contracts increases Gas costs, which may offset price improvements for small transactions. On Gas's low-cost Layer 2 network, this overhead is negligible. Compare the total costs (including Gas) of intent-based conversions and direct AMM transactions to see net benefits.

The future of intent-based transactions

Intent-based architecture is expanding beyond simple token redemption. By mid-2026, several trends have emerged.

Multi-action intent

Current intent describes a single action (exchanging token A for token B). The next generation system allows compound intents: "Exchange A for B, deposit B in the loan agreement, and lend C" as a single atomic intent. Solvers that can efficiently execute the entire sequence compete for the bundle.

AI-driven solvers

Machine learning models are being applied to solver optimization. The AI solver can predict short-term price changes, predict cross-chain liquidity conditions, and dynamically adjust routing strategies. The computational advantages of AI-driven solvers may accelerate the trend towards centralizing solvers because only well-resourced teams can train and operate these models.

Intent-Aware Wallet

Wallet began using intent-based execution by default for all redemption, making the intention layer invisible to the user. MetaMask integrates Uniswap's API for intra-wallet redemption, indicating that in the future, each wallet redemption will be automatically auctioned through the solver, and users will only see quotes and confirmations.

Regulatory Framework

As intent-based transactions grow, regulators are beginning to review whether solver activity constitutes regulated financial services. The pay-for-order flow (PFOF) model in traditional stock markets is structurally similar to the approach of some protocol compensation solvers. Regulatory clarity will determine whether intent-based transactions remain unlicensed or require licensed participants.

Frequently Asked Questions

What is the difference between a transaction and an intention?

The transaction specifies exactly how to perform an action: invoke this contract, use these parameters, and pay so many Gas. Intention specifies expected results: I want to exchange this for that, at least receive so much. The execution details are left to the solvers, who compete to find the best path.

Do I need to trust funds with the solver?

Not required. In a well-designed intent protocol, the solver never manages user funds. The user signs an intent authorizing the settlement contract to transfer tokens only when the solver delivers the promised output. Smart contracts enforce atomic conversions. If the solver cannot deliver, the transaction will not be executed.

How do solvers make money?

Solvers profit from the spread between the price at which they can gain liquidity and their bids in the auction. If a solver can buy 1 ETH from various sources at a price of 3,195 USDC and fill in the user's intentions at a price of 3,200 USDC, then the solver retains the spread of 5 USDC minus the Gas cost. Competition among solvers compresses this profit margin over time.

Can intentions be reviewed?

Intention submitted to a specific off-chain system of an agreement can theoretically be reviewed by the agreement operator. This is a different trust assumption than submitting transactions to the Ethereum decentralized memory pool. Some protocols alleviate this problem by running an open solver network that anyone can participate in.

Is intent-based redemption always cheaper than direct AMM transactions?

Normally yes, but not guaranteed. Intention based protocols provide better prices when the solver is highly competitive and there is enough transaction flow to achieve batch optimization or coincidental matching of needs. For extremely small transactions or during periods of low solver activity, the improvement may be minimal.

What happens if there is no solver to fill in my intentions?

This depends on the protocol. UniswapX falls back to the Uniswap route on the direct chain, so transactions are still executed at AMM prices. CoW Protocol expires unfilled orders after the batch window and users can resubmit them. Across intends to expire if no repeaters are filled within the deadline.

What are the intentions to prevent MEV?

The intention is a signed message submitted off-chain, not a transaction in a public memory pool. Because the MEV robot performs a pincer attack by monitoring pending transactions in the memory pool, removing transactions from the memory pool eliminates the attack vector. Solvers execute transactions on the chain, but solvers are mature actors that can protect against MEVs during their own execution.

Can I use intent to conduct cross-chain transactions?

Yes. UniswapX supports cross-chain intentions, where users can redeem tokens on one chain and receive tokens on another chain. Across Protocol is built entirely around cross-chain intentions for bridging. The solver handles cross-chain execution and verification results of settlement contracts on both chains.