ERC-AI is the token standard for Autonomous Agents in the Agentic ecosystem.
Where ERC-20 standardized fungible value, and ERC-721 standardized unique ownership, ERC-AI standardizes autonomous intelligence as an on-chain asset.
ERC-AI defines how an agent is identified, permissioned, upgraded, and integrated across networks.
ERC-AI is designed to enable:
Agents must be portable across products, chains, and execution layers.
Agents must operate under explicit permissions and constraints.
Agent origin, upgrades, and governance should be traceable.
Wallets, apps, and protocols should be able to integrate agents consistently.
Agents must be able to participate in incentives, fees, staking, and routing.
An ERC-AI token represents a persistent autonomous entity with:
Identity (who/what the agent is)
Capability profile (what it can do)
Permission scope (what it is allowed to do)
Upgrade path (how it evolves)
Integration hooks (how apps and protocols interact with it)
ERC-AI implementations typically include:
Agent Metadata
Role, creator, version, capability profile
Permissioning
Action scopes, signing policies, spending limits
Upgrade & Evolution
Versioned modules, capability extensions, governance-driven updates
Interaction Hooks
Standard events and callable methods for surfaces (e.g. Jam) and protocol components
Economic Hooks
Fee routing, staking participation, treasury controls, settlement endpoints
The goal is consistency: any ERC-AI-compatible agent can be recognized and integrated.
ERC-AI is a standard for agents to function as on-chain actors.
ERC-AI makes it possible to build:
Agent marketplaces
Agent reputations
Agent economies
Agent-to-Agent commerce
Institutional-grade autonomous execution flows
ERC-AI is presented in Agentic documentation as a protocol standard proposal.
Specification-level details (interfaces, events, and reference implementations) will be published as the protocol hardens.
For builders, the key point is simple:
ERC-AI is the foundation for a multi-chain agent economy.
Last updated