Ethereum Proposes Asset-Level Spending Limits for Wallets

Ethereum developers are discussing a new "asset-enforced spend mandate" proposal, posted to the Ethereum Magicians forum on June 18, 2026, that would let a token itself block a transfer, rather than relying only on wallet or session-key permissions, if it exceeds a per-transaction cap, has expired, was revoked, or involves a disallowed asset. The design specifically targets AI agents and delegated wallets that move funds without a human signing each transaction, giving a compromised or misbehaving agent's key a hard limit enforced at the token layer instead of relying on wallet-level controls alone. The proposal's author, Ethereum Magicians user VladKuzR, is now coordinating with the authors of the related ERC-8226 "Regulated Agent Mandate" standard and MetaMask's delegation-framework team to align the two approaches. It remains an early discussion draft, not a finalized ERC.
For teams building AI agents that move onchain assets, this proposal is worth tracking not because it's final, it isn't, but because it names a real gap precisely: wallet-level permissions do nothing once an agent's key is compromised, while a control baked into the token itself keeps working even then. Whether that lands as its own standard or folds into ERC-8226, the design pattern it's arguing for is becoming the expected baseline for agent wallet permissions.
What happened
An Ethereum Magicians forum post from June 18, 2026 proposes an "asset-enforced spend mandate," a draft standard that would let an ERC-20 token consult a small gate contract before allowing a transfer, rejecting it with a machine-readable reason (no mandate, expired, revoked, wrong asset, or over the per-transaction cap) if the transfer doesn't fit terms a token holder previously authorized. The design directly targets AI agents and delegated wallets moving funds without per-transaction human sign-off: because the constraint lives at the token contract rather than the wallet or session key, the proposal argues a transfer stays capped even if an agent's key is fully compromised.
Technical context
The proposal, from Ethereum Magicians user VladKuzR, positions itself as the missing layer in an existing standards stack: ERC-20 defines what you hold, ERC-3643/ERC-7943 define who may hold it, ERC-4337/ERC-7702 define who may sign for it, and this proposal would define what a holder may spend, enforced by the asset itself. Forum commentary flagged a real trade-off: token-level enforcement only works for tokens that opt in to the new interface, so it doesn't help existing ERC-20s unless each one upgrades. An alternative, account-level approach, ERC-7710/ERC-7715 and MetaMask's Delegation Framework, enforces spending limits before a transaction is even submitted and works with any existing token, but only catches transfers that route through the delegation system in the first place, not a holder who controls their own balance directly.
For practitioners
Developers building AI agents that move onchain assets should read this as a signal, not yet a standard to implement: per-transaction caps, expiration windows, allow-lists, and instant revocation are becoming the expected baseline for agent wallet permissions, whether enforced at the token layer, the account/delegation layer, or both. The two approaches are complementary rather than competing; asset-level gates matter most for regulated or issuer-controlled tokens where compliance must live in the transfer path, while account-level delegation is the more practical default for general-purpose agent spending across existing tokens.
What to watch
As of late June 2026, the proposal's author is coordinating with the authors of ERC-8226 (Regulated Agent Mandate), a closely related identity-and-compliance-focused standard, to see whether the two specifications should merge rather than compete; a working group and formal pull request are expected to follow. Worth tracking: whether the spend-mandate concept converges into ERC-8226 or ships as a separate standard, and whether any wallet providers or token issuers commit to implementing either approach.
Key Points
- 1A June 18, 2026 Ethereum Magicians proposal would let tokens themselves block transfers that exceed a spending mandate, targeting AI agents and delegated wallets.
- 2The design enforces caps, expiration, and revocation at the token layer, so a transfer stays limited even if an agent's key is fully compromised.
- 3The proposal's author is now coordinating with the related ERC-8226 standard's authors and MetaMask's delegation team to potentially merge the approaches.
Scoring Rationale
Strengthened from a single news write-up to three independently verified sources including the actual Ethereum Magicians primary document, showing the proposal is gaining real engagement from the ERC-8226 authors and MetaMask's delegation team. Still an early draft, not adopted, so kept in the notable-but-not-major tier.
Sources
Public references used for this report.
Practice with real Payments data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Payments problems
