Intent-Based Trading Explained

What is Intent-Based Trading?

Intent-based trading is a paradigm shift from transaction-based execution. Instead of submitting a signed transaction that specifies exactly what to execute, users express an intent: "swap 1 ETH for at least 1,500 USDC." The intent is submitted to a relayer or solver network, which competes to fill the intent profitably. Solvers aggregate intents, find optimal execution paths, and settle on-chain. Users never broadcast their full execution plan to mempool—MEV searchers can't sandwich them.

Intent-based trading is pioneered by UniswapX (Uniswap Labs), CowSwap (CoW Protocol), Across (protocol for cross-chain), and 1inch Fusion (DEX aggregator). The intent paradigm is formalized in ERC-7683 (Ethereum Execution Standard for Cross-Chain Intents). Adoption is growing as users demand MEV protection and better UX.

Intent Architecture & Flow

User Intent Submission

User creates an intent (e.g., ERC-712 signed message) specifying: (1) Input token & amount (1 ETH). (2) Output token & minimum amount (1,500 USDC). (3) Deadline & nonce (expiration time). (4) Optional: solver constraints, gas budget, slippage. Intent is NOT a full transaction—it's a statement of preference. User signs with private key but doesn't broadcast to public mempool.

Relayer/Solver Network

User submits intent to a relayer (centralized or decentralized). Relayer aggregates intents from many users, runs filtering/validation checks, and broadcasts to solver network. Solvers are MEV-aware operators (trading firms, AMM DEXs, market makers) who bid for the right to fill intents. Multiple solvers compete, ensuring best execution.

Solver Fulfillment & Auction

Solvers simulate execution: can they swap 1 ETH for 1,500 USDC? If yes, they bid (e.g., "will pay 2 gwei + $10 gas refund"). Best bid wins. Winning solver aggregates other intents, finds optimal execution path (Uniswap V3 + Curve + AMM combinations), and creates a settlement transaction.

On-Chain Settlement & Reconciliation

Winning solver submits settlement transaction on-chain. All user intents in the batch are executed atomically. Users receive outputs (1,500+ USDC) to their addresses. Solver captures remaining MEV (spreads) or covers any loss if execution was worse than bid. Settlement is final and immutable once on-chain.

Intent-Based Protocol Comparison

UniswapX Order Flow Auction

Permit2 & Order Signing

UniswapX uses Permit2 (ERC-2612 extension) to enable token approvals via signature. Users sign an order (ERC-712 struct) specifying: input token, output token, amount, slippage tolerance, deadline. Order is not a transaction—it's a promise: "I authorize this exact swap at this slippage." Orders are cryptographically signed but stay off-chain.

Order Flow Auction (OFA)

UniswapX Relayer collects orders, runs auction: fillers bid on the right to execute orders in exchange for spread profit. Example: user orders 1 ETH -> 1,500 USDC. Filler A bids 1,490 USDC (user gets promised min, filler keeps spread). Filler B bids 1,510 (user gets better). B wins. B executes the swap on-chain and pockets 10 USDC spread (or covers loss).

Filler (Solver) Incentives

Fillers profit from order flow, not MEV. They aggregate intents, find optimal DEX paths, and settle. Large fillers include: market makers (Wintermute, Amber Group), DEXs (Curve, Balancer), and MEV bots (redeployed for intent-based execution). Competition ensures low spreads for users. Fillers also pay gas, incentivizing batching (many intents per tx).

Real-World Example

You swap 1 ETH for 1,500 USDC on UniswapX. Relayer broadcasts order to 10+ fillers. Top 3 bids: A=1,490, B=1,520, C=1,515. B wins (best bid). B executes on-chain: swaps 1 ETH -> 1,520 USDC via Uniswap V3. You receive 1,520 USDC (5 USDC better than promised). B pockets the spread minus gas.

CowSwap Batch Auction Model

Coincidence of Wants (CoW)

CowSwap's unique feature: matching opposite orders directly. If you sell 1 ETH for USDC and another user buys 1 ETH with USDC, they can settle peer-to-peer with zero slippage, zero MEV. CoW reduces reliance on on-chain execution. Batch solver attempts CoW matching first, then falls back to on-chain DEX execution for unmatched orders.

Batch Auctions (15-Second Cycles)

CowSwap collects orders into batches every ~15 seconds on mainnet. Within each batch, solver finds optimal settlement: (1) Match CoWs directly. (2) Route remaining via DEXs. (3) Minimize total cost. Solver runs complex optimization off-chain. Settlement is atomic—all orders settle or none.

Solver Competition & SLA

Multiple solvers bid on batches. Each solver proposes settlement with a cost (gas + swap slippage). Lowest-cost solver's settlement is executed. CowSwap has Service Level Agreements (SLAs) ensuring minimum solver participation. If solvers fail to cover batch cost, fallback mechanism activates. Users always get execution.

Price Discovery & Fairness

CowSwap uses "clearing price"—a single price all matched orders settle at. This is fair and prevents MEV: both buyer and seller get same clearing price. If clearing price is worse than external market, order is rejected (too risky). CowSwap's fairness mechanism is stronger than UniswapX, but batch delays (15 sec) mean slower execution.

Solver Networks & Competition

Who Are Solvers?

Solvers are entities with capital, execution infrastructure, and arbitrage capabilities. Types: (1) Market makers (Wintermute, Amber Group): profit from spreads, use inventory. (2) Aggregators (1inch, Matcha): route through DEXs, capture comissions. (3) DEX operators (Uniswap, Curve): internalize liquidity. (4) MEV-repurposed bots: sophisticated operators adapting to intent model. Solvers need to bid competitively or lose order flow.

Solver Economics

Solvers profit from spread between order price and execution cost. Example: order 1 ETH -> 1,500 USDC, solver executes 1 ETH -> 1,512 USDC. Spread = 12 USDC minus gas. Gas cost on Ethereum: ~2-5 USDC. Net profit: 7-10 USDC per order. Volume-based: 1,000 orders/day = $7k-10k profit. Solvers must handle: capital requirements, execution risk, slippage, market risk.

Decentralized Solver Networks

Today's solvers are mostly centralized (companies). Future: decentralized solver networks where anyone can become solver by staking capital. EigenLayer restaking + intent-solving creates new opportunities. Decentralized solvers enable permissionless competition, reduce relayer power, improve censorship resistance. Early-stage protocols (Anoma, MEV-Burn) research this.

Solver Selection & Reputation

Relayers can whitelist solvers or accept open bidding. Metrics: (1) Historical execution quality (price improvement). (2) Reliability (settlements always succeed). (3) Gas efficiency. (4) Capital availability. Top-rated solvers get more order flow, can bid tighter. Poor performers are slashed or removed. Reputation is critical in solver competition.

MEV Protection Mechanisms

Off-Chain Order Collection

Intent-based systems don't broadcast orders to public mempool, eliminating mempool MEV (frontrunning, sandwich attacks). Users submit intents privately to relayers. Orders never reach public view until settlement. Searchers can't see and can't manipulate order flow.

Solver Bidding & Competition

Multiple solvers compete for orders. Solvers bid their execution quality (price improvement). Best bid wins. Competition prevents any single solver from extracting excess MEV. Users benefit from solver competition—prices improve as solvers bid higher. This is a form of MEV redistribution: MEV goes to users (via price improvement), not to validators.

Batch Auction & Clearing Prices

CowSwap's batch auction uses clearing prices: all orders in batch settle at same price. This prevents order-dependent MEV (where later orders see earlier price). All orders see identical pricing, ensuring fairness. Batch settles atomic, preventing partial fills that could create MEV opportunities.

Builder MEV & Future Defenses

Intent-based trading reduces application-level MEV but doesn't eliminate builder-level MEV (block proposers still reorder transactions). Future defenses: (1) Encrypted mempools (threshold encryption, TEE). (2) Threshold encryption (transactions encrypted until reveal). (3) MEV-Burn (burn MEV value instead of extracting). (4) Distributed builders (PBS, external builders reduce validator centralization).

FAQ