Author: imToken
Have you recently been overwhelmed by "crayfish" on your screen, or have you already tried it yourself?
This open-source agent, called OpenClaw (originally Clawdbot, later renamed MoltBot), has rapidly spread across the global tech community and crypto space in recent times, demonstrating a new emerging trend in AI agents: by leveraging open-source models, automation tools, and on-chain infrastructure, individuals for the first time have the opportunity to possess sustainable productivity units that operate independently of large platforms.
But behind this optimistic expectation, a more fundamental and thorny question is emerging: when AI agents begin to truly collaborate and start "hiring" each other, what gives us the right to trust them?
Against this backdrop, on January 30, Davide Crapis, head of AI at the Ethereum Foundation, tweeted that ERC-8004 has been launched on the Ethereum mainnet, and the team will deploy the singleton to all major L2s in the coming weeks.
This seemingly low-key technical advancement may have filled in a long-missing piece of infrastructure for the AI Agent economy.
I. What is ERC-8004?
If you still remember the dAI team at the Ethereum Foundation, it's not hard to understand the inevitability of the emergence of ERC-8004.
As previously mentioned, on September 15, 2025, the Ethereum Foundation established an artificial intelligence team called "dAI," with the core mission of allocating resources to define standards, incentives, and governance structures for AI models on the blockchain, including model credibility, that is, how to make AI behavior verifiable, traceable, and collaborative in a decentralized environment (further reading:Ethereum's "Second Curve": TradFi and AI Enter Simultaneously, a Trillion-Dollar Settlement Layer Beyond EVM Quietly Taking Shape）。
ERC-8004 is being promoted as one of the core standards in this context.
Unlike payment protocols such as x402, ERC-8004 does not directly address "how money is transferred," but instead attempts to answer a more fundamental question:How can AI Agents be identified, trusted, and participate in economic collaboration without platform endorsement?
This is precisely why the promotion team behind ERC-8004 is particularly impressive: led by the dAI team at the Ethereum Foundation, in collaboration with Google, Coinbase, and MetaMask in its formulation, it almost entirely covers the three key entry points of AI, transactions, and wallets. This level of involvement itself clearly signals that it is not just an application-layer experiment, but rather a long-term infrastructure bet.
As its official name is Trustless Agents, its core logic is not a complex algorithm, but aims toEnable AI to have verifiable identities, reputations, and capability proofs on the blockchain,In simple terms, its design is very restrained, doing only three things:
- Identity Registry:Based on the ERC-721 standard, each AI Agent will be "NFT-ized." This means that an AI Agent in a wallet is no longer just a line of code, but a unique ID-owning, transferable, and verifiable digital entity. It signifies that for the first time, an AI Agent is no longer merely an instance within a certain system, but can be looked up, referenced, and integrated into other protocols, just like a wallet address;
- Reputation Registry:It can be understood as the "Dianping" of the AI world, allowing users or other agents who have interacted with an agent to submit feedback. More importantly, these reviews can be linked to on-chain payments or custodial actions, ensuring that reputation is not a narrative generated out of thin air, but a historical record built upon real economic behavior. In an environment without centralized referees, this design, which grants evaluation rights only after payment, attempts to migrate the most fundamental credit logic from the real world into the AI collaboration network;
- Verify the registry:For high-value or high-risk tasks, historical reputation alone is not sufficient. ERC-8004 therefore reserves a third-party verification interface, allowing endorsement of an Agent's capabilities or execution process through trusted execution environments, zero-knowledge proofs, and other methods. Although this part has not yet been fully opened, its mere existence already indicates the standard's intent toward long-term complex collaboration scenarios;
It is worth noting that although born in the Ethereum ecosystem, the goal of ERC-8004 has never been Ethereum-specific from the beginning; instead, it was designed as a general standard for AI Agent discovery and trust through blockchain.
The standard is already being used or tested on multiple mainstream EVM networks, including Arbitrum, Base, Monad, and there is a clear plan to expand to some non-EVM ecosystems. This means that the ERC-8004 is not trying to solve a local demand of a specific chain, but rather the general trust issue that AI Agents face in cross-platform and cross-organizational collaboration.
2. What problems can ERC-8004 solve?
If you look back at the development path of AI over the past two years, you will find a very clear turning point.
In the stage where AI still mainly exists as a "tool," all problems can be attributed to model capabilities, computational costs, and product experience; but when AI Agents in 2025 become entities capable of independently accepting tasks, allocating resources, performing operations, and taking responsibility for the results, people suddenly realize that in this seemingly logical evolutionary process, a long-ignored issue has been infinitely magnified.
Why should AI agents trust each other?
You must know that a single agent cannot complete all tasks. It must improve efficiency by dividing labor, outsourcing, and reusing others' capabilities, just like in human society. This means that under this structure, AI must actively call upon other AIs to form a highly automated collaborative network.
Of course, in today's Web2 or platform-based AI applications, this issue doesn't exist, after all, trusting an AI essentially means trusting the platform, company, and brand behind it, and once a mistake occurs, the responsibility can be traced back to a centralized entity. But in an open, permissionless, decentralized Agent world, this logic no longer holds:An AI Agent may come from an individual, a DAO, a research institution, or even a completely anonymous deployer. It may perform well today but suddenly turn malicious tomorrow. It may claim to have certain capabilities, but you cannot verify its true origin.
This means that when AI agents begin to truly participate in economic activities, especially in high-risk behaviors such as asset management, transactions, settlements, and authorizations, "trust" will become a more scarce resource than "capability." Without an infrastructure that is queryable, verifiable, and accountable, large-scale collaboration among AI agents simply cannot be established.
It is precisely in this context that ERC-8004 was proposed, not to address what AI can do, but to directly tackle a more fundamental proposition: how AI can be treated as a "trustworthy entity" in a world without centralized guarantors. Therefore, in terms of design philosophy, ERC-8004 is not a complex technical breakthrough; it deliberately maintains extreme simplicity:
As mentioned above, under the framework of ERC-8004, each AI Agent is no longer just a vague program instance, but possesses a recognizable, referenceable on-chain identity. This identity is not intended for trading or speculation, but rather serves as a long-lasting anchor for binding capability claims, historical behaviors, and future responsibilities.
More importantly, this identity is not tied to a specific platform or application, but exists in an open public ledger, which means that anyone, any protocol, or any Agent can query and verify it under the same standard.
Beyond identity, ERC-8004 introduces an on-chain reputation system, which is particularly crucial because it attempts to address a long-standing issue that has plagued decentralized systems, namelyHow to build a credible historical record without platform arbitration.By binding evaluations to real tasks, actual payments, or custody actions, ERC-8004 aims to introduce the common sense of the real world—"you can only evaluate after using"—into the economic system of AI Agents. Reputation is no longer a marketing narrative, but a long-term accumulation of behavior.
When identity and reputation are still insufficient to cover high-value scenarios, ERC-8004 further reserves an interface for verification. This is not to prescribe a specific technical path, but to allow different forms of third-party verification mechanisms to be introduced, such as trusted execution environments, staking guarantees, or even zero-knowledge proofs.
Overall, ERC-8004 is not a standard for a single feature, but rather building a minimal viable "social structure" for AI agents, guiding AIs to start collaborating, competing, and taking responsibility like social members, establishing the most basic order.
Three, why Ethereum?
That would raise a new unavoidable question: in a future where AI agents are highly automated and frequently invoked, why would this set of standards choose Ethereum, instead of those faster, cheaper, or specifically designed for AI or high-frequency interactions blockchains?
The answer may not lie in the technical parameters themselves, but rather in an implicit asset accumulated by Ethereum over the long term: the credible neutrality as a global settlement layer.
It is important to emphasize that in the collaboration network of AI agents, what is truly expensive is not the communication cost, but the cost of errors. After all, once AI agents are allowed to directly operate assets, execute transactions, or act on behalf of others, any failure or malicious act could result in irreversible losses.
In such an environment, participants are undoubtedly more concerned about whether the rules are stable, whether records are tamper-proof, and whether responsibilities can be traced, and Ethereum has precisely established long-term advantages in these dimensions.
After all, Ethereum does not belong to any single company or consortium. Its security model, audit culture, and ecosystem maturity have already been repeatedly validated in DeFi, NFTs, and institutional-level applications. Therefore, the proposal of ERC-8004 is precisely an attempt to extend these existing advantages to AI Agents, a completely new type of entity.
Looking further,Ethereum is competing here for a more subtle new label: the underlying settlement layer for AI collaborationIn this vision, AI agents from different platforms, different chains, and different organizations can operate within their respective environments, but when they need to establish trust, custody value, or settle outcomes, they will ultimately return to a neutral layer that is mutually recognized.
It can be said that this new role aligns highly with Ethereum's gradually established position in the global financial system: it does not strive to become the fastest execution layer, nor does it attempt to cover all application scenarios, but instead aims to occupy a more fundamental position in the long term.The credible settlement and order layer.
Of course, ERC-8004 will not immediately give rise to explosive applications; it addresses a problem that has not yet fully arrived. But from a longer time perspective, when AI agents evolve from mere "functional modules" into autonomous "economic entities" capable of independently accepting orders, collaborating, and settling transactions, and when wallets no longer serve only human individuals but also begin to carry the authority boundaries, liability attribution, and risk isolation of AI, this foundational order regarding identity, reputation, and verification may very well become a prerequisite for whether the AI economy can truly scale.
In this sense, ERC-8004 is not an attempt at a short-term narrative, but rather Ethereum's clear bet on its own future: in an economic system where both humans and AI participate.Order and settlement still require a trustworthy, long-term neutral foundation.
Although at this moment when the bear market is roaring and the mood is low, people may not be very interested in Ethereum's technical upgrades, and it's even hard to maintain enough patience.
But there is no denying that as new production relations are quietly taking shape, Ethereum is striving to capture this new direction that has not yet fully emerged but is sufficient to reshape the future.