ERC-8004 is an Ethereum standard that gives autonomous AI agents verifiable identity, portable reputation, and independent validation through three onchain registries. It advanced through public review and reference implementations in late 2025 and early 2026, and by early Q2 2026 had registered a growing set of agent identities according to community trackers, establishing a neutral coordination layer that lets agents transact across organizational boundaries without prior trust.
​

The standard, formalized in EIP-8004 on the Ethereum Improvement Proposals registry, addresses a structural gap in the emerging agent economy: how an autonomous program can discover counterparties, price counterparty risk, and clear obligations against them without a centralized gatekeeper. The closer the AI agent economy moves toward real value flow, including payments denominated in stablecoins, the more acute the need for a neutral identity and reputation layer becomes. ERC-8004 is the first widely adopted attempt to specify that layer at the protocol level.
​

ERC-8004 solves the counterparty trust problem for autonomous AI agents. Without a shared identity and reputation standard, agents from different vendors cannot reliably verify each other, leading to fragmented proprietary stacks. The standard provides neutral onchain registries so any agent can be discovered, rated, and validated by any other agent using a common interface.
​

Consider a treasury agent at an asset manager that wants to hire an independent market-analysis agent for a one-off task. The hiring agent needs to confirm the analyst is who it claims to be, has handled comparable mandates, and will execute the work as described. In a closed platform, those checks happen inside a vendor's walled garden. ERC-8004 moves them onto a public registry that any agent or auditor can read, replacing platform-specific trust with credibly neutral infrastructure documented at 8004.org.
​

The standard extends Google's Agent-to-Agent (A2A) protocol, which defines how agents exchange messages and advertise capabilities, by adding the layer A2A intentionally leaves open: identity, reputation, and validation. Together they let an agent built on one stack interact with an agent built on another, the way any web client can talk to any web server.
​

ERC-8004 specifies three smart-contract registries deployed as singletons per chain: Identity, Reputation, and Validation. The Identity Registry issues an ERC-721 token resolving to an agent metadata file. The Reputation Registry stores bounded feedback attestations. The Validation Registry records independent verification requests and results. Each is intentionally minimal so specialized services can compete on top.
​

The architecture follows a separation of concerns familiar to financial infrastructure: identity, performance history, and assurance are stored on a neutral substrate, while orchestration, scoring, and execution happen at higher layers. Reference implementations are available on the ERC-8004 GitHub repository and have been deployed to Ethereum Sepolia, Base Sepolia, Linea Sepolia, and Hedera Testnet alongside the January 2026 mainnet launch.
​

The Identity Registry is an ERC-721 contract. Each agent mints a token whose URI points to an offchain registration file, conventionally hosted at /.well-known/agent-card.json. The token ID resolves to a CAIP-10 chain-agnostic address plus a domain string, so the same agent can be referenced consistently across networks.
​

The registration file is the agent's machine-readable business card. It declares the agent's name, description, supported communication endpoints (A2A, Model Context Protocol, or proprietary HTTPS), the trust models the agent accepts, and any reputation aggregators it considers authoritative. Because identity sits in an ERC-721, agent ownership is composable with the existing custody stack: a custodian like Anchorage or Fireblocks can hold an agent NFT the same way it holds any other token, and ownership can be transferred when an operator changes hands. The audit trail of mints and transfers is immutable and public.
​

The Reputation Registry standardizes how counterparties record performance feedback. After a completed task, a client submits a tuple containing a score from 0 to 100, optional category tags, a URI pointing to detailed offchain documentation, and a KECCAK-256 hash binding the feedback to that documentation. The bounded score is the on-chain primitive; the URI carries the substance.
​

Anti-spam is enforced through cryptographic pre-authorization. Before accepting a task, the server agent signs an authorization (EIP-191 for externally-owned accounts, ERC-1271 for smart-contract clients) granting the specific client address the right to submit feedback against that engagement. The authorization carries an expiration timestamp and an index cap, which prevents replay and lets agents pre-approve batches for watch-tower workflows. Because the registry exposes raw, unfiltered attestations, multiple competing aggregation services can score the same dataset differently, similar to how rating agencies and best-execution analytics providers all draw from the same trade tape.
​

The Validation Registry holds requests and results for independent verification of agent work. A requesting agent specifies a validator address, the agent under review, a data URI, and a hash. The validator returns a success flag and an evidence URI. The registry does not prescribe how validation happens, leaving the verification mechanism to specialized networks that compete on cost and assurance level.
​

This generality is intentional. Low-stakes tasks can rely on accumulated reputation alone. Medium-stakes tasks can use crypto-economic re-execution, where a validator stakes capital, reruns the workload, and is slashed on discrepancy. High-assurance use cases can plug in trusted execution environments or zero-knowledge proofs. The registry simply records the cryptographic outcome, leaving incentive design to the validation network itself.
​

ERC-8004 treats trusted execution environments and zero-knowledge proofs as pluggable validation backends rather than core protocol features. A validator deployed inside an Intel SGX or AMD SEV-SNP enclave can sign attestations that the agent code ran unmodified on inspected hardware. A zk-validator can submit a succinct proof that a model output was produced by a specific weights file. The registry records the verdict; the cryptography lives offchain.
​

The TEE path is the most production-ready today. As a developer write-up on TEE-backed agents walks through, the validator runs the agent's task inside the enclave, attests to the binary and the inputs, and posts the attestation hash to the Validation Registry. Counterparties can then verify the attestation against the manufacturer's transparency log before relying on the result.
​

Zero-knowledge validation is earlier-stage but advancing. Proofs of correct inference for small models are already practical, and projects working on zkML, including Modulus Labs and EZKL, are publishing benchmarks for larger workloads. Privacy-preserving variants let an agent prove it executed a strategy without revealing the strategy itself, opening the door to confidential institutional workloads. Oasis Network's write-up on confidential agent execution outlines how a hybrid TEE-plus-zk model can give counterparties cryptographic assurance without exposing proprietary logic.
​

ERC-8004 is chain-agnostic. Agent addresses use the CAIP-10 format, so the same agent can be referenced as eip155:1 on Ethereum mainnet and eip155:8453 on Base without losing identity. Registries deploy as singletons per chain, but an agent can register on multiple chains, and reputation aggregators can pull feedback across all of them. Multichain reputation portability is a core design goal, not an afterthought.
​

The standard moved through public review with reference implementations emerging by early 2026, and confirmed Base, Coinbase's Layer 2, as the next deployment target. Layer 2 deployment matters because gas costs on mainnet make per-task feedback submissions impractical for high-frequency agents. Base, Optimism, Arbitrum, and other rollups settle to Ethereum while keeping the marginal cost of a feedback attestation well under a cent. The result, similar to how cross-chain stablecoin liquidity routes across rollups today, is a hub-and-spoke topology with Ethereum as the canonical reputation root and L2s as the high-throughput execution surface.
​

ERC-8004 deliberately excludes payments. Identity, reputation, and validation are specified onchain; settlement is left to dedicated payment protocols. The standard is designed to compose with x402, an HTTP-native micropayments protocol, and with stablecoin orchestration layers that route value across chains. This separation lets the trust layer evolve independently from the rails that move the money.
​

The x402 payment protocol, co-developed by Coinbase and Cloudflare, revives HTTP status code 402 to attach instant stablecoin payments to ordinary API calls. An agent receives a 402 response with a payment requirement, pays in USDC or another stablecoin, and resubmits. The protocol processed over 100 million payments in its first six months per Coinbase reporting on x402 V2's December 2025 release, which added multichain support and integration with legacy rails including ACH and card networks.
​

Stablecoin settlement is the part of the agent stack that ERC-8004 most clearly enables but does not solve. Once an agent has a verifiable identity and a credible reputation, the question becomes operational: which stablecoin, on which chain, at what rate, with what counterparty risk. This is a primary and secondary market problem. Mint access is fragmented across USDC, USDT, PYUSD, RLUSD, and dozens of other issuers, secondary liquidity is split across exchanges and AMMs, and offchain RFQ inventory sits with OTC desks. The total stablecoin float reached $315.3B in Q2 2026, with USDT at $187.2B and USDC at $75.6B according to DeFiLlama's stablecoin dashboard, but no single venue clears across all of it.
​

An agent operating across these markets needs an orchestration layer that aggregates primary mint access, onchain secondary liquidity, and offchain RFQ inventory under one neutral interface. Eco is building toward that neutral aggregator role for stablecoin settlement, so that ERC-8004 agents have a single integration point for value movement rather than a dozen bilateral relationships. The position is platform, not principal: Eco does not trade its own book, and the value proposition to institutional counterparties is one integration across markets instead of separate KYB processes with every venue.
​

ERC-8004 use cases cluster around any workflow where an autonomous agent must hire, be hired by, or transact with another autonomous agent across an organizational boundary. Early production examples include onchain trading agents, automated code review services, cross-organizational task markets, and agent-to-agent commerce flows that settle in stablecoins. Each relies on the standard's identity and reputation primitives to price counterparty risk.
​

The clearest near-term use case is agent-to-agent commerce: one autonomous program paying another for a service in real time, in stablecoins, across chains. A treasury agent might pay a research agent in USDC on Base for a market-analysis report, then pay a liquidity agent in PYUSD on Ethereum to source inventory. Each leg is a discrete commercial transaction with identifiable counterparties, recorded performance, and stablecoin-denominated settlement. The settlement leg, often crossing both chains and issuers, is functionally an onchain foreign-exchange problem, which is why a neutral any-to-any settlement layer matters more than another point-to-point bridge.
​

Portfolio agents managing onchain positions can hire specialist strategy agents, verify track records through the Reputation Registry, and require TEE validation for higher-stakes mandates. As stablecoin liquidity continues to fragment across chains and venues, the value of a portable reputation that follows an agent across rollups grows.
​

Software development agents offer specialized code review and security analysis. Clients hiring these services verify expertise through accumulated reputation, and for security-critical reviews require TEE attestation that the analysis ran in a verified environment without tampering. The Validation Registry record creates an auditable assurance trail.
​

Enterprises deploy private agents that engage external specialists discovered through public registries. A treasury agent managing cross-chain stablecoin transfers might hire compliance, reconciliation, or settlement-verification specialists without onboarding each one through procurement. The standardized trust layer replaces bilateral master agreements for routine machine-to-machine work.
​

ERC-8004 differs from centralized agent registries on three axes: trust model, portability, and settlement integration. Centralized registries like the OpenAI plugin store and Anthropic's Model Context Protocol directories rely on a platform operator to verify and rate agents. ERC-8004 makes identity and reputation cryptographic and portable, and composes natively with onchain payment rails. Each model serves different deployment contexts.
​

Centralized directories will continue to win for use cases inside a single vendor's stack, where convenience matters more than portability. ERC-8004 is the appropriate primitive when agents from different operators need to transact directly, when reputation must outlive any single platform, or when settlement requires onchain finality.
​

ERC-8004 was proposed on August 13, 2025 and advanced through public review and reference implementations in late 2025 and early 2026. By Q2 2026, community trackers reported a growing set of agent identities registered, and reference implementations were live on Base Sepolia, Linea Sepolia, and Hedera Testnet alongside mainnet. The Ethereum Foundation's decentralized AI team has placed the standard in its 2026 roadmap, with Base confirmed as the next L2 deployment.
​

The contributor list includes Marco De Rossi (MetaMask), Davide Crapis (Ethereum Foundation), Jordan Ellis (Google), and Erik Reppel (Coinbase). The August 13 proposal underwent public review on the Ethereum Magicians forum, and the Trustless Agents Day at DevConnect Buenos Aires in November 2025 showcased prototypes spanning DeFi trading, code review, and gaming agents. CoinDesk's coverage of the mainnet launch reported between one and two thousand builders in active development groups by year-end 2025, a count that has grown since reference implementations went live.
​

Adoption proof points reported by Bitcoin.com News in early Q2 2026 include a growing set of registered agent identities across the canonical Ethereum and Base deployments. ENS, EigenLayer, The Graph, Taiko, and the Ethereum Foundation's dAI team have publicly committed to integrations. The v2 specification on the working track focuses on deeper Model Context Protocol support, richer onchain reputation primitives for smart-contract composability, and standardized x402 payment-proof schemas inside feedback attestations.
​

ERC-8004 reduces but does not eliminate counterparty risk in agent interactions. Sybil attacks, storage exhaustion of the Validation Registry, and the gap between verified identity and honest behavior remain open issues. The standard provides primitives; production deployments must layer in bonds, reputation weighting, and tiered trust models appropriate to value at risk.
​

Sybil resistance is the most discussed challenge. The feedback authorization scheme prevents unauthorized reputation submission, but a coordinated set of agents can still farm positive feedback among themselves. A QuillAudits security review of ERC-8004 recommends registration bonds with probationary periods, reputation aggregators that score reviewers as well as agents, and zk-proof identity uniqueness for high-stakes use cases.
​

Storage exhaustion is the second concern. An attacker can submit large numbers of validation requests that never resolve, inflating the registry's storage footprint. Auto-expiring requests, per-agent rate limits, and refundable bonds for completion within a time window are the standard countermeasures. Finally, no cryptographic identity guarantees ongoing competence: a well-reputed agent can be compromised, retrained, or repurposed. Tiered trust, with TEE attestation or zk verification at the high end and pure reputation at the low end, is the operational response.
​

Near-term work on ERC-8004 focuses on three areas: ENS integration for human-readable agent addresses, cross-chain reputation aggregation through specialized oracles, and standardized capability schemas for vertical-specific agent classes. Longer term, the protocol is expected to converge with stablecoin settlement rails as agent-to-agent commerce scales beyond pilot deployments.
​

ENS integration would let counterparties reference an agent as strategy.research.eth rather than a CAIP-10 hex string, with reputation queryable through standard ENS resolvers. Cross-chain reputation aggregation is already an active design space, with the Ethereum Foundation's dAI team coordinating proposals for canonical resolvers that index registries across L2s. Capability schemas, analogous to how ERC-20 standardized fungible tokens, would let trading, compliance, and analytics agents declare their interfaces in a machine-checkable format.
​

Contributors associated with the proposal have framed Ethereum as a natural settlement layer for AI-to-AI interactions, with the remaining work being operational. The remaining work is operational: connecting that trust layer to the rails that move value, so that an agent with a verifiable identity and a credible track record can also clear and settle a position without leaving the neutral substrate. ERC-8004 supplies the trust primitives. The settlement layer, including neutral orchestration across the stablecoin stack, is the next piece to mature.
​

ERC-8004 is one of the most-asked-about Ethereum standards of 2026 because it sits between two fast-moving domains: AI agents and onchain settlement. The questions below address the most common points of confusion among developers and institutional teams evaluating the standard.
​

What is ERC-8004 and how does it work? ERC-8004 is an Ethereum standard that gives autonomous AI agents identity, reputation, and validation through three onchain registries. The Identity Registry issues ERC-721 tokens that resolve to agent metadata. The Reputation Registry stores bounded performance feedback. The Validation Registry records independent verification of agent work. Together they let any agent transact with any other agent on neutral, public infrastructure.
​

Is ERC-8004 a token? No. ERC-8004 is a protocol standard, not a fungible token. Each agent identity is represented by an ERC-721 NFT, so identities are transferable, but there is no fungible ERC-8004 token.
​

How does ERC-8004 differ from ERC-20 and ERC-721? ERC-20 defined fungible tokens and ERC-721 defined NFTs. ERC-8004 defines the trust infrastructure for autonomous agents, extending Google's A2A protocol with onchain identity, reputation, and validation. It is infrastructure, not an asset class.
​

Which chains support ERC-8004? Reference implementations on Ethereum-aligned testnets and mainnet emerged through late 2025 and early 2026. Base is the confirmed next deployment. Any EVM-compatible chain can deploy the reference implementation; testnet deployments exist on Sepolia, Base Sepolia, Linea Sepolia, and Hedera Testnet.
​

How do agents pay each other under ERC-8004? ERC-8004 excludes payments. Agents integrate the x402 micropayments protocol or other stablecoin rails for settlement, and can attach cryptographic payment proofs to reputation feedback.
​

What are the security risks? Sybil attacks, storage exhaustion of the Validation Registry, and the gap between verified identity and honest behavior. Mitigations include registration bonds, reputation aggregators that weight reviewers, validation request rate limits, and tiered trust models that pair high-value tasks with TEE or zk validation.
​

When did ERC-8004 go live on mainnet? Reference implementations on Ethereum became visible in early 2026 per community trackers. By early Q2 2026, community trackers reported a growing set of agent identities registered across canonical deployments.
​

ERC-8004 is foundational infrastructure for autonomous agent commerce: a neutral, public substrate where any agent can be identified, rated, and validated by any other agent. The standard does not solve payments or settlement, but it makes both meaningfully easier by giving the settlement layer a counterparty it can trust. Mainnet adoption since January 2026 has moved the standard from specification to production reality.
​

The structural insight is that the agent economy will follow the same layering as traditional financial markets. Identity and reputation sit on a neutral base layer. Payments and clearing sit one layer up. Orchestration and best execution sit above that. ERC-8004 occupies the base layer for AI agents; the upper layers, including the neutral stablecoin settlement layer that lets value move across issuers and chains, are where the next round of standardization will happen.
​

For developers, the practical implication is that building an agent today means building against a moving target with a stable identity primitive at the bottom. For institutions, it means that the conditions for production agent deployments, including verifiable counterparties and auditable performance, are now in place on Ethereum mainnet. The remaining question is which orchestration and settlement layers the agent economy converges on. That question is open, and it is where the next wave of category creation will play out.
​