A new Ethereum standard called ERC-8211 would let AI agents chain multi-step DeFi transactions into a single atomic call, resolving each step's parameters at execution time rather than requiring them to be fixed at signing.
Biconomy has proposed ERC-8211, an Ethereum standard that lets AI agents and smart accounts chain multi-step DeFi transactions into a single atomic call — resolving each step's parameters at execution time rather than requiring them to be fixed when the transaction is signed.
The proposal, published on 6 April and now open for comment on the Ethereum Magicians forum, addresses a problem that has become acute as AI-driven trading volumes grow. Most existing batch systems require every parameter to be set before a transaction hits the chain. That works for simple transfers. It falls apart when step two depends on the output of step one — the exact proceeds of a token swap, the precise collateral ratio after a deposit, the actual amount returned from a lending withdrawal. In those cases, the signer must either guess the output value and risk the transaction reverting, or split the operation into multiple transactions and pay gas for each one while exposing themselves to front-running between steps.
ERC-8211's "smart batching" approach solves this with a contract-layer encoding where every input parameter carries three pieces of information: a fetcher type that defines how the value is sourced — as a literal, via a static call, or from an on-chain balance; routing information that decides whether it becomes a call target, value field, or calldata; and inline predicates that must hold true or the entire batch reverts.
The result is that an AI agent can express a flow like "swap USDC for ETH on Uniswap, then deposit whatever actually arrives into Aave" as a single transaction, with the second step dynamically pulling its amount from the resolved output of the first call rather than a guessed number.
The standard works with existing account-abstraction frameworks — specifically ERC-4337 smart accounts — and requires no changes to the Ethereum protocol itself. That's a deliberate design choice. Proposals that require hard forks face years-long adoption timelines; ERC-8211 can be deployed by any smart account provider today.
The Ethereum Foundation has indicated the standard aligns with its "Improve UX" strategic priority, which aims to hide DeFi complexity from end users. Foundation support matters because ERC proposals without it tend to languish in committee indefinitely. With it, ERC-8211 has a realistic path toward broad adoption — though "realistic" in Ethereum governance terms still means months of review, reference implementations, and audits before anything touches production.
The practical significance extends well beyond AI agents, though that's where the demand is most urgent. Autonomous trading systems — the kind that now account for an estimated 60 to 65 per cent of crypto trading volume — currently break complex strategies into sequential transactions, each of which leaks value to MEV bots and incurs separate gas costs. A single atomic batch with runtime resolution eliminates both problems. Early benchmarks cited in the proposal suggest gas cost reductions of up to 89 per cent for multi-step DeFi operations, though real-world figures will depend on the specific flow and network conditions.
Biconomy isn't a neutral party here. The company built its business on account-abstraction infrastructure; its SDK is already embedded in hundreds of DeFi applications. ERC-8211 would make Biconomy's existing tooling more valuable by giving it a standardised batch format to build around. That commercial interest doesn't invalidate the technical merits — the proposal is well-engineered and solves a genuine problem — but it does explain why Biconomy is the one writing it rather than, say, the Aave team or Uniswap Labs.
The timing coincides with a broader shift in how DeFi protocols think about composability. The explosion of AI agents operating on-chain has exposed the limitations of transaction models designed for human users who click buttons in a web interface. An AI agent doesn't need a confirmation modal or a three-step approval flow; it needs an expressive, atomic instruction set that can encode arbitrarily complex financial operations. ERC-8211 is the first serious attempt to standardise one.
Whether it becomes the standard depends on competing proposals, audit results, and whether major wallet providers choose to implement it. The Ethereum Magicians discussion is already active, with contributors from Safe, Alchemy, and several MEV research groups weighing in. For now, ERC-8211 is a proposal — nothing more. But it's a proposal that solves a problem the market has been working around rather than addressing head-on, and it arrives at the exact moment when the workarounds have become untenable.
