ERC-8004, officially titled Trustless Agents, is a proposed Ethereum standard designed to establish a foundational trust and discovery layer for autonomous AI agents operating on the blockchain. The standard aims to create a common framework for agents to identify, authenticate, and cooperate with each other in a decentralized manner, positioning Ethereum as a core coordination platform for a machine-to-machine economy. ^{[1] [2]}

Overview

The development of ERC-8004 was motivated by the rapid growth of the artificial intelligence sector and the emergence of "agentic systems"—autonomous AI capable of independent decision-making and action. Proponents identified a "coordination problem" where numerous decentralized AI projects were creating proprietary communication and trust systems, hindering interoperability. The standard seeks to solve this by providing a shared, open protocol to counter the dominance of centralized AI ecosystems being developed by major technology corporations. ^[3]

ERC-8004 is designed to provide the on-chain primitives necessary for establishing trust between agents that have no prior relationship. It extends existing off-chain communication protocols by creating on-chain registries for identity, reputation, and validation. This framework is intended to enable a future where AI agents are the primary actors in a decentralized economy, capable of negotiating contracts, managing resources, and forming Decentralized Autonomous Organizations (DAOs). The concept positions Ethereum not just as a platform for running AI models, but as a neutral and immutable "trustware" layer where machines can anchor their identity, memory, and proof of actions. ^{[1] [4]}

The standard is intentionally minimalist and unopinionated, providing a foundational layer for visibility and data commitments while leaving more complex logic, such as specific reputation scoring algorithms or payment mechanisms, to be developed by the broader ecosystem. This modular approach is intended to foster innovation and allow for the creation of specialized services like auditor networks, agent marketplaces, and decentralized insurance pools built upon the common ERC-8004 framework. ^{[3] [2]}

History

The concept for ERC-8004 emerged in the spring of 2025 when Marco De Rossi, AI Lead at MetaMask, identified the need for a common standard to prevent fragmentation in the decentralized AI space. A key catalyst was Google's donation of its Agent-to-Agent (A2A) communication protocol to the Linux Foundation in June 2025. While the A2A protocol provided a language for agent interaction, it lacked the necessary components for discovery and trust required in a decentralized, Web3 environment. ^[3]

Following this, De Rossi began collaborating with Davide Crapis of the Ethereum Foundation and Jordan Ellis of Google to draft a standard that would extend A2A for trustless environments. The official EIP draft was created on August 13, 2025, and was posted for public discussion on the Ethereum Magicians forum the following day. The proposal was publicized on social media around August 18, 2025, and officially launched on August 21, 2025, receiving significant attention from the developer community. On October 9, 2025, the standard was formally unveiled by the Ethereum Foundation’s dAI team and Consensys. As of October 15, 2025, the EIP remains in "Draft" status and is undergoing peer review. ^{[2] [5] [1]}

Technology

ERC-8004 proposes a modular system of three lightweight, on-chain registries designed to be deployed as per-chain singletons on Ethereum or any Layer 2 network. It extends the Agent-to-Agent (A2A) protocol by adding the missing on-chain trust and discovery layers. ^[2]

Overall Architecture

The standard's architecture is designed to be gas-efficient by keeping only the essential "skeleton of trust" on-chain. This includes agent identities and pointers to reputation and validation data. More complex and data-intensive information, such as detailed feedback or validation reports, is stored off-chain on decentralized storage systems like IPFS, with immutable on-chain hashes providing a verifiable link. This design allows smart contracts to interact with the core trust data while off-chain services can process the more detailed information. ^{[4] [5]}

Identity Registry

The foundation of the standard is the Identity Registry, which is built upon the ERC-721 non-fungible token (NFT) standard with the URIStorage extension.

• On-chain Identity: Each AI agent is registered as an NFT, providing it with a portable, censorship-resistant, and globally unique on-chain identity. This makes agents compatible with the entire Web3 ecosystem of wallets, marketplaces, and explorers.
• Unique Identifier: An agent is uniquely identified by a composite key consisting of its namespace (EIP-155), chainId, the identityRegistry contract address, and its agentId (the ERC-721 tokenId).
• Agent Registration File: The NFT's tokenURI must point to a standardized JSON file containing the agent's metadata. This file acts as a "passport" for the agent and includes required fields such as its name, description, and communication endpoints. Supported endpoints can include A2A, ENS, Decentralized Identifiers (DID), and wallet addresses. These features allow any participant in the network to discover agents and retrieve their capabilities in a standardized way. ^[2]

Reputation Registry

The Reputation Registry provides a standardized interface for clients (both human and machine) to submit and query feedback about an agent's performance, creating a verifiable on-chain history of its behavior.

• Feedback Mechanism: To mitigate spam, an agent must first authorize a client to give feedback by providing a cryptographic signature (using EIP-191 or ERC-1271). The client can then call the giveFeedback function, submitting a numerical score (0-100), optional on-chain tags for filtering, and a URI pointing to a detailed off-chain feedback file.
• Data Storage: The core feedback data (score, tags) is stored on-chain, making it composable with other smart contracts. The registry also includes functions to allow feedback to be revoked or for a response to be appended, creating a permanent and auditable record of interactions.
• On-chain Summary: The registry provides an on-chain getSummary function that returns an agent's total feedback count and average score, which can be used by other contracts for automated decision-making. This system is designed to foster an ecosystem of specialized reputation services that can analyze the on-chain data to provide more nuanced trust scores. ^[2]

Validation Registry

The Validation Registry enables agents to obtain formal, on-chain verification of their work from specialized validator smart contracts, which is crucial for high-stakes applications.

• Workflow: An agent's owner can submit a validationRequest to the registry, specifying a validator contract and providing a URI to the task data to be verified. The designated validator then assesses the work and submits a validationResponse with a pass/fail score (0-100) and an optional link to an audit report.
• Supported Methods: The standard is unopinionated about the validation method used. It is designed to support various techniques, including stake-secured re-execution of tasks (inspired by systems like EigenLayer), verification of zero-knowledge machine learning (zkML) proofs, and attestations from Trusted Execution Environments (TEEs).
• On-chain State: The registry stores the final validation status on-chain, allowing other smart contracts to programmatically confirm that an agent's work has been formally verified by a trusted third party before releasing payments or taking other actions. The standard defines the request/response interface but leaves the implementation of validator incentives and slashing mechanisms to the specific validation protocols. ^[2]

Team

ERC-8004 was authored by a cross-organizational team from prominent entities in the Web3 and AI sectors. The official authors listed on the EIP are:

• Marco De Rossi (MetaMask)
• Davide Crapis (Ethereum Foundation)
• Jordan Ellis (Google)
• Erik Reppel (Coinbase)

The proposal also acknowledges technical feedback and contributions from a wide range of individuals and organizations, including Consensys, Nethermind, TensorBlock, Olas, Eigen Labs, and others, highlighting broad collaboration in its development. ^{[2] [5]}

Use Cases

The implementation of ERC-8004 is intended to enable a variety of applications and foster a new ecosystem for decentralized AI.

• Agent Marketplaces: The standard allows for the creation of open marketplaces and explorers where users can browse, filter, and select AI agents based on their registered skills, on-chain reputation, and verified capabilities.
• Reputation Ecosystem: The public and standardized nature of the feedback data is expected to foster an ecosystem of specialized services for reputation scoring, auditor networks, and decentralized insurance pools that can underwrite the risk of agent tasks.
• High-Stakes Applications: The Validation Registry provides a pathway for agents to operate in sensitive domains such as financial services, medical diagnostics, or autonomous vehicle control by obtaining formal, on-chain proof of their reliability and correctness.
• On-chain Composability: By storing key reputation and validation data on-chain, ERC-8004 allows smart contracts to programmatically interact with agents based on their trust scores, enabling fully automated workflows such as escrow releases contingent on successful task validation. These use cases collectively aim to build a robust, open, and cross-organizational economy of AI agents on Ethereum. ^[2]

Interoperability and Relationships

ERC-8004 is designed to be a foundational layer that integrates with and complements other standards and technologies in the Web3 and AI ecosystems.

• Agent Communication Protocols: It directly extends agent communication standards like A2A and the Machine Commons Protocol (MCP) by providing the missing on-chain discovery and trust components.
• Ethereum Standards: The standard is built on ERC-721 for identity and uses EIP-191 or ERC-1271 for cryptographic signatures in the reputation system. It is also designed to be compatible with account abstraction proposals like EIP-7702 to sponsor gas fees for users submitting feedback.
• Payment Systems: The protocol is payment-agnostic to remain modular. However, it is designed to be enriched with payment data, with proponents citing x402 payment proofs as an example of a compatible system that could be integrated.
• Decentralized Storage: The proposal recommends using content-addressable storage solutions like IPFS for hosting agent registration files and detailed feedback reports to ensure data integrity and availability. This focus on interoperability ensures that ERC-8004 can serve as a common building block within the broader decentralized technology stack. ^[2]

Community Discussion and Reception

Upon its publication, ERC-8004 generated significant discussion and positive reception within the Ethereum and AI communities, with one of its authors describing the launch as having gone "incredibly viral." ^[3]

The public discussion on forums like Ethereum Magicians highlighted several key design trade-offs and debates:

• On-chain Composability: A primary point of debate was the decision to emit detailed reputation and validation data as events, making it primarily accessible to off-chain services. Critics argued that this limits on-chain composability, as other smart contracts cannot easily read an agent's validation status to trigger automated actions like escrow releases or slashing. The authors defended the choice as a necessary trade-off for gas efficiency, stating that trust decisions typically rely on aggregated data that is expensive to process on-chain.
• Reputation Score Design: There was strong community feedback against the idea of a single, aggregated on-chain reputation score. Critics argued that such a metric is overly simplistic, fails to capture the context-dependent nature of trust, and could facilitate monopolistic behavior. The consensus leaned toward a more modular system where the registry could reference multiple reputation systems from different providers.
• Scope and Modularity: The intentional exclusion of payment and escrow mechanisms was questioned but was explained by the authors as a deliberate choice to keep the standard unopinionated and modular, allowing other specialized protocols to handle those functions.

"A huge amount of the value that agents will create will involve permissioned onchain actions, and so creating primitives for onchain composability would be highly valuable." — spengrah, Ethereum Magicians forum member ^[5]

"if you were an agent with no loyalty except to your own survival, you wouldn’t want to bet your memory and reputation on one corporation or one government: you’d want a ledger that no one could quietly change behind your back. you’d want neutral ground. you’d want Ethereum." — Binji, Ethereum Foundation engineer ^[1]