ERC-8004: Ethereum's Answer to the AI Agent Trust Problem
When AI agents start transacting with each other across organizational boundaries, one question becomes critical: how do you trust an agent you've never met?
ERC-8004, the "Trustless Agents" standard proposed by a coalition from MetaMask, the Ethereum Foundation, Google, and Coinbase, offers a compelling answer. It's not just a technical specification—it's an attempt to build the trust infrastructure that an open agent economy demands.
And from a Good Agents perspective, it gets a lot right.
What ERC-8004 Actually Does
The standard introduces three lightweight on-chain registries:
Identity Registry — Every agent gets a unique, portable identifier (based on ERC-721). Think of it as DNS for the machine economy. An agent's identity resolves to a registration file describing its capabilities, endpoints, and supported protocols like A2A or MCP.
Reputation Registry — A standard interface for posting and retrieving feedback signals. Crucially, only authorized parties can post feedback about an agent—you can't inflate your own reputation or trash a competitor's.
Validation Registry — Hooks for independent verification: stakers re-running jobs, zkML verifiers, TEE oracles, or trusted judges. This is where cryptographic proof meets real-world accountability.
Payments are deliberately left out of the spec. The standard integrates with protocols like x402 (Coinbase/Cloudflare's HTTP 402 revival) but doesn't prescribe how agents should pay each other. This is wise—payments are complex, and keeping the trust layer minimal means it can adapt as payment rails evolve.
Where ERC-8004 Aligns with FAR-SIGHT
Let's evaluate ERC-8004 against the Good Agent principles:
Transparent ✓ — This is ERC-8004's strongest suit. On-chain identity and reputation create an immutable, auditable record. An agent's history is public. You can see who validated their work, what feedback they received, and whether they've been flagged. No black boxes.
Governed ✓ — The Validation Registry provides hooks for compliance checking, independent audits, and cryptographic verification. Agents operating under ERC-8004 can be subjected to governance frameworks—regulatory requirements, organizational policies, or community standards—with on-chain evidence of compliance.
Integrated ✓ — The standard is explicitly designed for interoperability. It extends Google's A2A protocol, supports MCP endpoints, and works with ENS names and DIDs. An ERC-8004 agent can advertise multiple endpoints and prove control of domains. This is infrastructure for a multi-protocol agent ecosystem.
Sustainable ✓ — By standardizing identity and reputation at the protocol level, ERC-8004 reduces the coordination costs that would otherwise fragment the agent economy into proprietary silos. Agents built on this standard aren't locked into a single vendor's ecosystem.
Human-focused ? — This is where it gets interesting. ERC-8004 is primarily agent-to-agent infrastructure. Human oversight isn't baked into the protocol—it's left to implementations. A Good Agent built on ERC-8004 would need additional mechanisms to ensure humans remain in the loop for consequential decisions.
Accurate ? — The standard provides verification hooks but doesn't guarantee accuracy. A validated agent isn't necessarily an accurate agent. The Reputation Registry records feedback, but feedback is subjective. Accuracy testing remains the responsibility of agent developers and the validators they choose.
Resilient ? — ERC-8004 addresses some resilience concerns (portable identity, decentralized reputation) but introduces others. What happens when a validator network fails? When reputation data becomes stale? When an agent's keys are compromised? The standard includes identity recovery mechanisms, but resilience at scale remains unproven.
Fruitful ? — The financial value proposition depends entirely on what gets built on top. ERC-8004 is infrastructure. It reduces trust friction, which should reduce transaction costs and enable new agent-to-agent markets. But the ROI accrues to the applications, not the protocol itself.
The Real Innovation: Deliberate Minimalism
What makes ERC-8004 interesting isn't just what it includes—it's what it deliberately excludes. The standard is lean. It doesn't specify how reputation should be scored, how validators should be compensated, or how disputes should be resolved. These are left to "off-chain components" and specialized services.
This minimalism is strategic. A maximalist standard would be brittle—it would encode today's assumptions and break when those assumptions change. By keeping the on-chain footprint small, ERC-8004 creates space for experimentation. Different reputation algorithms can compete. Different validation mechanisms can emerge. The market can discover what works.
For Good Agents practitioners, this means ERC-8004 is a foundation, not a complete solution. You still need to build the governance frameworks, the accuracy testing, the human oversight mechanisms, and the resilience engineering. The standard gives you portable identity and verifiable reputation—what you do with them is up to you.
Risks to Watch
A few concerns worth flagging:
Reputation bootstrap problem — New agents start with no history. How do they compete with established agents? Staking and delegated reputation are proposed solutions, but they favour well-capitalized incumbents.
Sybil attacks — Creating multiple identities to manipulate reputation remains possible. The feedback authorization mechanism helps, but determined attackers can still create networks of colluding agents.
Concentration risk — Early movers with strong reputations could dominate, recreating the centralization the standard aims to prevent. The Ethereum Foundation has included reputation decay mechanisms, but network effects are powerful.
Complexity cost — Adding blockchain infrastructure to agent systems increases complexity. For many use cases, traditional trust mechanisms (contracts, SLAs, vendor relationships) may remain simpler and cheaper.
The Bottom Line
ERC-8004 represents serious thinking about how to build trust infrastructure for autonomous agents. It aligns well with several Good Agent principles—particularly Transparency, Governance, and Integration—while leaving others to implementation.
If you're building agents that need to operate across organizational boundaries, transact with unknown counterparties, or build verifiable track records, ERC-8004 is worth understanding. It's not a complete governance solution, but it's a credible foundation for one.
The standard is currently in draft status. The Ethereum Foundation's dAI team is actively promoting it, and reference implementations are deployed on multiple testnets. Whether it achieves the adoption of ERC-20 or ERC-721 remains to be seen—but the eight independent implementations within 24 hours of release suggest the market sees value here.
For Good Agents, the question isn't whether to adopt ERC-8004. It's whether your agents need the kind of cross-boundary trust that ERC-8004 enables—and if so, what additional layers you'll need to build on top to achieve genuine FAR-SIGHT compliance.