Across Protocol pioneered the intent-based bridging model in 2022 and has since become one of the largest solver networks in production. Its architecture has three components: a single-sided liquidity hub on Ethereum, a competitive solver (called "relayer" in Across terminology) network, and an optimistic verification system that releases funds to relayers after a dispute window. The design has filled $14B+ in volume with median fill time of 8 seconds per Across stats as of April 2026.
​

Understanding Across matters for two reasons. First, it is one of the reference implementations of ERC-7683, having migrated its production order format in Q3 2025. Second, its solver-relayer model is the template that newer intent networks have adopted with variations. Eco integrates with Across-style solver networks as one of the primitives in its 15-chain stablecoin orchestration layer.
​

Across is an intent-based bridging protocol with a hub-and-spoke topology. The hub is on Ethereum mainnet; spokes are deployed on every supported destination chain (currently 15+ including Base, Arbitrum, Optimism, Polygon, BNB, Linea, Blast, Mode, Scroll, ZKsync, and Lisk per Across docs). Users deposit on the source spoke; relayers deliver on the destination spoke; the hub reconciles the deposit and pays the relayer.
​

The protocol launched in November 2022 with $50M in initial liquidity from Risk Labs (the development entity behind UMA, which provides Across's optimistic oracle). By April 2026, Across had processed 12.4M cross-chain transfers per its Dune dashboard, with a median user-perceived completion time of 8 seconds and 99.7% successful fills.
​

The user flow has four stages. First, the user calls deposit() on the source spoke pool, locking the input asset and emitting a deposit event with the destination chain, recipient, and minimum output amount. Second, off-chain relayers monitor the source spoke's deposit events; the first relayer to commit on the destination chain wins the order. Third, the relayer calls fillRelay() on the destination spoke, transferring the output asset directly from the relayer's inventory to the recipient. Fourth, after a dispute window (typically 1-2 hours), the relayer claims reimbursement from the hub via the optimistic oracle.
​

The two-stage design — fast user delivery, slow relayer reimbursement — is what makes Across feel instant to users while preserving capital security. The user receives funds in seconds because the relayer pays from inventory. The relayer waits hours to be paid back because the hub uses optimistic verification: a dispute window allows challengers to flag fraudulent fills, and only after no challenge can the relayer claim. Across's intent-architecture documentation describes the full state machine.
​

Across calls its solvers "relayers." A relayer is an off-chain agent that holds inventory on every chain it serves, monitors source-chain deposit events, and commits to fills profitable to the relayer's price model. As of April 2026, Across has 40+ active relayers per its public stats, with the top 5 handling roughly 70% of volume.
​

Relayer economics are spread-based. The user pays the input amount; the relayer delivers the minimum output amount; the difference (less the protocol fee) is the relayer's gross profit. Out of that profit, the relayer pays destination-chain gas, source-chain rebalancing costs (when inventory drifts), and capital cost. Median relayer take on stablecoin pairs is 4-7 basis points per Dune dashboards.
​

Relayer competition is thin in the first second after deposit (most fills are won by one relayer based on websocket monitoring speed), then widens for delayed orders. Relayers can submit bonds via Across's relayer registry, but participation is currently open without a strict bond requirement; the optimistic verification is the safety net.
​

Across uses UMA's optimistic oracle for cross-chain settlement. After a relayer fills on the destination chain, the relayer submits a reimbursement claim to the hub on Ethereum. The claim sits in a dispute window (default 2 hours, configurable per request). If no challenger disputes, the hub releases funds to the relayer. If a challenger disputes, UMA's oracle adjudicates by referencing the truth of the destination-chain fill event.
​

This is materially different from the validator-set model used by Hyperlane, LayerZero, or Wormhole. Across does not require a validator to sign every fill. It assumes most fills are honest, allows anyone to challenge, and bonds challengers and disputers to align incentives. The cost is the latency on relayer reimbursement (hours) versus the latency on validator-attested bridges (minutes). The benefit is no validator-set capture risk.
​

UMA's documentation on the cross-chain oracle describes the dispute mechanics in detail. As of April 2026, no fraudulent fill has been successfully disputed in production — relayers have strong economic disincentives to submit invalid claims because the bond is slashed if a challenge succeeds.
​

The hub is a single set of contracts on Ethereum mainnet that manages the protocol's accounting. Every relayer reimbursement settles against the hub. The spokes are lighter contracts on each destination chain that handle deposits (when the chain is the source) and fills (when the chain is the destination). The spoke contracts are upgrade-controlled by the hub via cross-chain messages.
​

This topology centralizes the accounting layer and decentralizes the execution layer. All financial state ultimately reconciles on Ethereum, which gives Across a single source of truth and makes the protocol's solvency auditable from one chain. Spokes are stateless aside from their per-chain inventory; they execute the protocol's logic without holding decision authority.
​

The trade-off is mainnet dependency. Every reimbursement claim consumes Ethereum gas, which means relayer profits are sensitive to mainnet gas conditions. Etherscan's gas tracker shows the cost variance: a relayer claim during a 30-gwei period costs ~$2; the same claim at 200 gwei costs ~$13. Across's relayers price this variance into their spread quotes.
​

Across, UniswapX, and CoW Protocol use different solver competition models. Across uses first-to-commit: the first relayer to submit a fill on the destination chain wins the order. UniswapX uses Dutch-auction decay: the order's acceptable price worsens over time, and solvers compete on when to accept. CoW Protocol uses batched auctions: solvers compete to find the best multi-order match within a 30-second batch.
​

Each model has consequences. First-to-commit (Across) rewards latency optimization — solvers race to be fastest. Dutch auctions (UniswapX) reward price patience — solvers wait for the order to decay to a profitable level. Batched auctions (CoW) reward optimization — solvers find clever multi-order matches. The result is different solver economics in each network and different fill-quality characteristics for users. CoW's protocol documentation compares the three models.
​

Across migrated to ERC-7683 as its production order format in Q3 2025. Per its public dashboard, ERC-7683 orders now make up 88% of Across volume. The migration replaced Across's proprietary deposit format with the standard GaslessCrossChainOrder struct, which means any wallet supporting ERC-7683 can submit Across orders without integrating an Across-specific SDK.
​

The orderData field encodes Across-specific routing parameters: the destination spoke address, the relayer reward tier, and the dispute-window duration. A solver network that supports the standard but not Across-specific routing can ignore the orderData and decline; a relayer that does support Across decodes the field and bids on the order. The architecture preserves Across's hub-and-spoke topology while making the user-facing order format portable.
​

Across's production metrics as of April 2026 (per its public stats and Dune analytics):

The successful-fill rate is high because Across allows users to specify minimum-output guarantees and relayers must meet them; failed fills typically result from order expiration when the relayer's price model rejected the spread, not from execution failure.
​

Across's optimistic settlement adds latency for relayers, which means relayer capital is tied up for hours per order. This concentrates relayer participation among well-capitalized actors who can afford the float. Smaller relayers struggle to participate at scale, which contributes to the top-5 concentration of volume.
​

The hub-on-Ethereum design means every reimbursement settles on Ethereum mainnet, which exposes Across to mainnet gas conditions. During high-gas periods, relayer reimbursement claims become expensive, which compresses relayer margins and can widen user-facing spreads. Etherscan's gas tracker shows mainnet gas spiked to 200+ gwei multiple times in Q1 2026, each spike correlating with a brief widening of Across spreads.
​

Across's coverage is EVM-only. The protocol does not natively serve Solana, Tron, Sui, or other non-EVM chains. Production teams needing cross-VM coverage layer Across underneath an orchestration layer that handles non-EVM via separate primitives.
​

Eco operates as an orchestration layer above multiple solver networks and bridge primitives, including Across-style intent networks. For an EVM-to-EVM stablecoin transfer, Eco's routing engine evaluates whether an Across-style fill is the cheapest and fastest path; for non-EVM destinations or for very large transfers, Eco routes through alternative primitives like CCTP or Hyperlane-secured solver inventory.
​

This is the orchestration-layer thesis: production teams want one integration that covers all routes, and the orchestration layer picks the right underlying primitive at runtime. Eco Routes architecture documentation describes the routing decision tree. The team integrates the Eco SDK once; the SDK handles solver selection, settlement, and finality across 15 chains.
​

Across has been in production since November 2022, which makes it one of the longest-running solver networks in the intent ecosystem. The operational history reveals several lessons that apply to the broader category.
​

Lesson one: relayer concentration is a feature in early-stage networks and a risk in mature ones. In the first year, having a small number of well-capitalized relayers ensured fills happened reliably. As the network matured, the same concentration created throughput bottlenecks during peak demand. Across has gradually opened relayer participation to smaller actors, but the top-5 concentration persists per Dune analytics.
​

Lesson two: optimistic settlement works in production but requires patient relayer capital. Across's two-hour dispute window means relayers float the destination-chain delivery for hours before reimbursement. This concentrates participation among well-capitalized actors and creates a barrier to entry for smaller relayers. The trade-off (no validator-set capture risk) is worth it for many use cases, but the capital implications shape the relayer market.
​

Lesson three: chain coverage is the dominant competitive surface. Across's expansion from 4 chains in 2022 to 15+ chains in 2026 directly drove volume growth. Chains added include Base (which became the largest single-chain volume contributor by mid-2025), Linea, Blast, Mode, Scroll, ZKsync, and Lisk. Each chain addition unlocked new flows and required relayers to expand inventory.
​

Yes. Across's hub-and-spoke contracts are deployed on Ethereum and EVM L2s. Non-EVM coverage requires layering Across underneath an orchestration layer that handles cross-VM transfers via separate primitives.
​

Across is an intent-based bridge with a relayer network. LayerZero and Wormhole are message-passing protocols. Across uses an optimistic oracle (UMA) for settlement; LayerZero and Wormhole use validator sets. The two layers are complementary — an intent-based bridge can use a message-passing protocol for verification.
​

A relayer is Across's term for what the broader intent ecosystem calls a "solver." It is an off-chain agent that holds cross-chain inventory and delivers destination-chain assets to fill user orders.
​

The dispute window is 2 hours by default, configurable per order. Relayers can request faster reimbursement by accepting a fee discount, or slower reimbursement at a fee premium. Capital efficiency tilts toward larger relayers who can absorb the float.
​

Across is primarily an asset-transfer primitive. The protocol does support a callable parameter that lets the destination-chain delivery trigger a contract call, but the canonical Across use case is asset bridging, not arbitrary cross-chain messaging. For richer cross-chain function-call patterns, apps typically use a dedicated message-passing protocol like Hyperlane or LayerZero alongside Across, or layer Across underneath an orchestration layer that handles both asset transfer and message-passing routing requirements together.
​