Summary: ERC-8118 is best understood as a capability-delegation layer for autonomous agents rather than as an identity registry, smart wallet, or generic signature primitive. Its core move is to let a principal authorize an agent address for specific function selectors with optional time windows and call-count limits, while also requiring agent-side consent and preserving the principal’s existing EOA custody. The useful mechanism lens is that ERC-8118 tries to standardize a persistent but bounded form of agent authority that sits between one-shot signatures and full smart-account migration.
What it does:
Defines a standard interface for principals to authorize agents for specific contract function selectors rather than broad wallet-wide power
Supports optional start and end timestamps plus usage-limited authorizations with automatic exhaustion after a configured call count
Requires EIP-712 signatures from agents as consent to being bound, which is meant to stop malicious principals from attaching agents to unwanted responsibilities
Enforces a single-principal constraint per agent per contract so one agent is not simultaneously bound to conflicting principals in the same authorization surface
Includes revoke, batch-authorize, batch-revoke, and view functions so contracts can treat agent permissions as a first-class on-chain capability map
Positions itself against ERC-4337 and EIP-7702 as a lower-migration way to get persistent, fine-grained delegated execution while keeping EOA custody intact
Key claims:
The abstract says ERC-8118 defines a standard interface for authorizing autonomous agents to perform specific on-chain actions on behalf of principals, which is the clearest reason to catalog it as delegated-execution authority infrastructure
The draft repeatedly emphasizes time-bounded authorizations, usage-limited authorizations, and function-level granularity, showing that its main design goal is bounded capability delegation rather than generic wallet abstraction
The specification requires agent consent via EIP-712 signatures, which is analytically important because the standard treats the agent as a consenting counterparty, not just a passive delegate target
The single-principal constraint suggests the design is optimizing for unambiguous accountability and simpler revocation semantics, even at the cost of making shared-agent business models harder
The rationale explicitly compares ERC-8118 with ERC-4337 and EIP-7702, arguing that it preserves EOA custody while offering more persistent and granular delegation than transaction-scoped code attachment
The draft’s use cases span automated trading, DeFi position management, DAO proposal execution, gaming NPCs, subscriptions, and cross-chain relayers, which signals that the authors see it as a reusable authority primitive rather than an application-specific scheme
The discussion thread and patch text both highlight production validation claims tied to prior deployment scale, but those claims should still be treated as draft-era author assertions rather than as standard-level proof of adoption
Whitepaper: No standalone ERC-8118 whitepaper or litepaper surfaced in this pass. The strongest primary materials were the Ethereum Magicians discussion thread and the draft ERC pull request / patch; see ../../whitepapers/erc-8118-primary-sources-2026-05-08.md.