MEV · Flashbots, MEV-Boost, SUAVE, order-flow auctions
ConfidenceLikelyUpdated2026-07-29Review by2026-10-27Sources9Machine-translatedOriginal (JA)
On this page
Wiki route
This entry sits under systems index. Read it against EigenLayer for the broader crypto-economic-coordination context, L2 agent economics for the rollup-level MEV picture, and agent-economy index for the agent-transaction-cost angle.
Key facts
- MEV (Maximal / Miner Extractable Value) = value validators / sequencers can extract by reordering, including, or excluding transactions
- Flashbots (2020) introduced a private-transaction relay (MEV-Geth) and later MEV-Boost (post-Merge), splitting block-building from block-proposing on Ethereum
- MEV-Boost adoption post-Ethereum Merge (2022-09): >90% of Ethereum blocks now constructed via MEV-Boost (a form of out-of-protocol PBS — Proposer-Builder Separation)
- SUAVE (Single Unifying Auction for Value Expression) is Flashbots’ next-gen architecture for cross-chain MEV / order-flow markets
- Order-flow auctions (OFAs) route user transactions to private channels where searchers compete to provide best execution; CoW Swap, UniswapX, 1inch Fusion are production examples
- For AI agents: MEV economics directly affect agent execution cost, slippage, and front-running risk
MEV taxonomy
| MEV type | Mechanism | Example |
|---|---|---|
| DEX arbitrage | Cross-pool price difference | Uniswap ↔ Sushi spread, cross-DEX cyclic arb |
| Liquidation | Trigger underwater positions | Aave / Compound liquidation bots |
| Sandwich | Front-run + back-run user swap | Pumping then dumping around victim trade |
| NFT sniping | Front-run mint / sale | Detecting underpriced listings |
| Backrunning | Insert after a target tx without front-running | Less adversarial; often considered “good MEV” |
| JIT (Just-In-Time) liquidity | Add liquidity exactly before a large swap | Uniswap V3 / V4 specific |
| Time-bandit / reorg | Re-mine blocks to capture missed MEV | Largely mitigated post-Merge on Ethereum |
Pre-Merge → Post-Merge architecture
Pre-Merge (PoW):
User tx ──► mempool ──► miner (chooses ordering) ──► block
│
MEV extracted directly
by miners or via Flashbots bundles
Post-Merge (PoS) with MEV-Boost:
User tx ──► mempool (public OR private OFA)
│
▼
Builders compete to construct most profitable block
│
▼
MEV-Boost relay (e.g., Flashbots, BloXroute, Eden)
│
▼
Validator selects highest-paying block
│
▼
Validator proposes; block finalized
Key shift: proposer (validator) is separated from builder. Validators no longer need MEV-extraction expertise; sophisticated builders compete to construct the most profitable block and bid for inclusion.
MEV-Boost numbers and dependencies
- >90% of Ethereum blocks built via MEV-Boost relays since 2023 ; major builders: beaverbuild, rsync-builder, Titan, Builder0x69, others
- Major relays: Flashbots, BloXroute, Eden, Ultra Sound, Aestus
- Censorship concern: some OFAC-compliant relays filter sanctioned addresses; Ethereum community pushed back via censorship-resistant relay diversification
This creates an economic dependency on relay operators and builders that resembles the dependency model in ERC-4337 bundlers and CCTP attestation — a centralization vector inside a nominally decentralized stack.
SUAVE — what it is trying to be
SUAVE (Single Unifying Auction for Value Expression) is Flashbots’ design for a decentralized block-builder network that:
- Operates as its own chain (SUAVE-chain) coordinating MEV across multiple destination chains
- Lets users / agents express preferences (“execute swap with max 1% slippage, route through best venue across all chains”)
- Searchers and builders compete to fulfill preferences off the destination chain, settle on-chain
- Targets cross-chain MEV unification — currently fragmented across Ethereum L1, every L2, Solana, Cosmos, etc.
Status (2025-2026): SUAVE testnet live; production rollout in progress. Strategic bet: as cross-chain volume grows and rollups proliferate, fragmentation creates inefficiency that SUAVE-style unification can capture.
Order-Flow Auctions (OFAs) for users / agents
OFAs route user transactions to a private auction where searchers compete to give the user the best execution:
| OFA / channel | Publicly documented mechanism | Agent integration boundary |
|---|---|---|
| CoW Protocol | users sign intents; solvers compete in batch auctions and may use coincidence of wants | verify order validity, signing domain, solver result and settlement contract |
| UniswapX | Permit2-based signed orders are filled by competing fillers under the order’s auction parameters | validate chain, nonce, deadline, reactor and output constraints |
| 1inch Fusion | signed swap intent is resolved by eligible resolvers through the Fusion auction flow | verify resolver / API terms, auction parameters, allowance and settlement status |
| MEV Blocker | an RPC endpoint routes transactions away from the public mempool and uses a searcher / backrun flow | RPC availability and policy are an added dependency; signed transaction limits remain the user’s responsibility |
Sources: ^[https://docs.cow.fi/] ^[https://docs.uniswap.org/contracts/uniswapx/overview] ^[https://help.1inch.io/en/articles/6800254-1inch-fusion-mode] ^[https://docs.mevblocker.io/]
For AI agents executing on-chain trades, OFAs offer structural MEV protection — instead of broadcasting to public mempool (sandwich risk), the agent signs an intent and a resolver competes to fill it.
L2 / rollup MEV implications
On L2 rollups (Arbitrum, Optimism, Base, etc.), the sequencer is the MEV extractor by default — a centralized sequencer can extract all MEV that flows through. Decentralizing sequencers is an active research area:
- Espresso, Astria — shared-sequencer networks (some atop EigenLayer AVS)
- PEPC / based rollups — alternative architectures where L1 validators sequence
- L2 OFAs — per-rollup OFA deployments to channel MEV transparently
For AI-agent volume forecasting on Base / Arbitrum / Optimism, MEV economics affect agent unit costs directly — sequencer revenue partly funded by MEV extraction is reflected in user fees.
Related
- INDEX
- eigenlayer-overview
- l2-agent-economics-arbitrum-base-op-comparison
- erc-4337-userop-bundler-flow
- rollup-market-share-2026-arbitrum-optimism-base
- solana-firedancer-validator-economics
- sui-aptos-move-l1-ecosystem
- bft-validator-economy-overview
- INDEX
- stablecoin-routing-agent-transactions
- FinWiki index
Sources
- flashbots.net for Flashbots project pages, MEV-Boost stats, and SUAVE documentation.
- writings.flashbots.net for research posts including PBS history.
- github.com/flashbots for MEV-Boost and SUAVE reference implementations.
- ethresear.ch for MEV-related research threads.
- Public MEV-Boost dashboards (e.g., mevboost.org-class trackers).
Discovery
Keep reading
Read next
- Pectra EIP-7251 · institutional staking economic fit (2048 ETH cap)The Ethereum consensus protocol was originally designed with a max 32 ETH per validator — this was for "decentralization first" and to let home users participate with 32 ETH. The cost was th...
- Ethereum Pectra Upgrade · 4 EIP Dual-Track Strategy OverviewPectra modifies the execution layer and consensus layer at the same time. EIP-7702 lets 9000 万 existing EOAs obtain batch / gasless / session key capabilities without changing addresses —— a...