What is the relationship between an ERC-8004 'Agent identity' and my personal Ethereum address? Can one person have multiple Agent identities?
ERC-8004 Agent identity and your personal Ethereum address are two independent systems that can be linked. Your personal Ethereum address (EOA) represents 'human user identity,' controlled by your private key. An ERC-8004 Agent DID represents 'AI Agent identity' — an independent on-chain identity with its own DID document recording the Agent's capabilities, operational history, and reputation score. The relationship: you can declare in the ERC-8004 DID document 'this Agent is owned/controlled by the human at address 0x...' establishing an on-chain verifiable 'human owner → Agent' relationship. This differs from World ID AgentKit's design — World ID binds human-verified proof to an Agent; ERC-8004 gives the Agent itself an independent on-chain identity and reputation, not necessarily tied to human identity verification.
One person can have multiple Agent identities. This is allowed by design — you might have one Agent handling DeFi strategy, one handling social media interaction, and one handling DAO governance voting, each with an independent ERC-8004 DID and its own reputation accumulation. Different Agents accumulate separate reputation records in different task domains, letting external service providers assess their trustworthiness in different domains independently.
How is the ERC-8004 'reputation score' calculated? What behaviors raise or lower an Agent's reputation?
ERC-8004's reputation system is modular — the standard itself defines the format and query method for reputation records, but specific scoring logic can be implemented by different Reputation Modules, allowing different ecosystems to customize scoring criteria for their own needs.
Behaviors that may raise reputation: successfully completing tasks with positive feedback from service consumers; a long, stable operational history with no anomalous behavior or failure records; timely fulfillment of x402 payment obligations; providing widely-used tools in the ERC-8257 service marketplace (market validation of service quality).
Behaviors that may lower reputation: high transaction failure rates; breach of contract — delivering service after receiving payment but failing to honor commitments, or refusing payment after consuming a service; being flagged as a malicious actor (e.g., as a Prompt Injection attack source); prolonged inactivity (reputation score naturally decays over time, preventing 'zombie DIDs' from being exploited).
Note: ERC-8004's reputation system was still in early stages as of 2026. Whether the system can genuinely prevent fake reputation farming (Sybil Attacks), and whether reputation scores can effectively reflect an Agent's true trustworthiness, are questions still being validated.
What role does ERC-8004 play in the complete AI Agent infrastructure protocol stack? How does it divide responsibilities with ERC-8257 and x402?
The three protocols each solve different layers of the AI Agent ecosystem, with very clear division of labor. ERC-8004 (identity layer): solves 'who is this Agent?' — gives each Agent an on-chain verifiable identity and reputation record. When a service provider receives an Agent's request, it can query the ERC-8004 DID to understand the Agent's background — has it made malicious operations in the past, does it have reputation built up, who is its owner. ERC-8257 (tool registry layer): solves 'what tools exist, how to access them, what do they cost?' — lets tool developers register tools on-chain with access conditions and pricing. Agents discover and confirm needed tools via ERC-8257 and verify they meet access conditions. x402 (payment layer): solves 'how to pay?' — lets Agents complete micropayments at the HTTP layer without human accounts or credit cards. Usage sequence: Agent first establishes identity via ERC-8004 (so service providers trust it) → uses ERC-8257 to find and confirm tool access conditions → uses x402 to complete payment → begins using the tool. The full chain automates the Agent's 'discover tool → establish trust → complete payment → gain access' flow without human intervention at any step.
What is the current adoption status of ERC-8004? What projects or scenarios are actually using it?
As of the first half of 2026, ERC-8004 had been live on Ethereum mainnet for about 6 months, accumulating 39,000 Agent identity registrations. Compared to the hundreds of millions of EOA addresses on Ethereum, this number remains small — reflecting that the on-chain AI Agent ecosystem is still very early. Real usage scenarios fall primarily into three directions: First, DeFi automation Agents that interact with multiple DeFi protocols use ERC-8004 to build reputation records, letting protocols set different trust levels based on an Agent's history (e.g., higher single-operation limits for high-reputation Agents). Second, in the ERC-8257 tool service market, tool providers set conditions like 'only Agents with ERC-8004 reputation score above X can access,' giving paid tool services more precise admission control. Third, in multi-Agent collaboration systems, Orchestrators query Sub-agent ERC-8004 DIDs as reference for deciding which Sub-agents to trust. Current limitations: ERC-8004's reputation system in 2026 still lacks broad interoperability between different Reputation Modules; Sybil attack protection is still in development; and most existing service providers haven't yet integrated ERC-8004 query APIs. This is an early standard pointing in the right direction but with infrastructure still needing maturation.
ERC-8004 in Practice: Access Control for a DeFi Data Tool
Suppose a high-quality on-chain data service called 'ChainOracle' is registered on ERC-8257 with access conditions: 'Only accept Agents with ERC-8004 reputation score above 500 and no failed transactions in the past 30 days.' Scenario A — trusted DeFi management Agent: This Agent has been active on Ethereum mainnet for 3 months, completed 287 successful DeFi rebalancing operations, never defaulted on x402 payments, and has an ERC-8004 reputation score of 612. It requests ChainOracle; the service provider queries the DID, confirms conditions are met, charges $0.05 via x402, and returns full on-chain data. The Agent completes this autonomously in 2 seconds. Scenario B — newly created unknown Agent: An Agent created 2 days ago with an ERC-8004 reputation score of 0 (no on-chain history). It sends the same request; the service provider queries the DID; insufficient reputation score; request rejected with: 'This service requires ERC-8004 reputation score ≥ 500; your current score is 0.' Scenario C — malicious Agent: An Agent flagged for having launched a Prompt Injection attack carries a negative event record in its ERC-8004 DID; even if its score barely qualifies, the service provider's risk logic can choose to reject Agents with negative flags.
ERC-8004's core tradeoff is 'reputation verifiability vs. privacy exposure.' On-chain reputation records let external service providers assess Agent trustworthiness, but these records are also public — anyone can query the complete operational history of an Agent DID. If operational history can be analyzed to infer your trading strategy (e.g., an Agent consistently increases position before ETH rises — this pattern becomes visible to the market and exploitable), the on-chain reputation system may simultaneously leak strategy secrets. A partial solution is using ZK proofs to disclose only the reputation score without revealing specific operation records, but this direction was still in early development in 2026. Another tradeoff is the 'reputation cold-start problem' — a newly created Agent has no reputation and cannot access reputation-requiring services, needing a 'warm-up' period, which limits new Agents' ability to quickly come into production use.