Can ERC-8004 replicate the success of x402 and become the cornerstone of the AI agent trust layer?

Original author: David, Deep Tide TechFlow

Original title: Taking over x402? It's time to pay attention to ERC-8004.

x402 is obviously popular.

According to CoinmarketCap data, the trading volume of various projects in the x402 ecosystem surged by 137 times, with the first ecosystem token PING reaching a market value of 30 million dollars from zero in just a few days.

Various KOLs have been intensively posting analyses, covering everything from technical principles to project reviews; every angle you can think of has been written about.

Two weeks ago, when we analyzed x402 earlier and mentioned the potential of projects like PayAI, there was not much buzz in the entire market.

In various narratives and the rapidly shortening versions of token lifecycles, researching new narratives in advance makes it easier to identify opportunities for related assets.

(Related reading: Google and Visa are both positioning themselves, what investment opportunities are hidden in the underestimated x402 protocol?)

Now, every time you refresh Twitter, a new “x402 ecological project” pops up; to be frank, if you are just starting to research x402 now, it might already be a bit late.

It's not that the protocol itself has no prospects, but rather that the most obvious Alpha opportunities have been fully explored.

But just when everyone was focused on x402, the observant would notice that another protocol has recently been frequently mentioned in discussions within the English crypto community:

ERC-8004.

Interestingly, one of the proposers of ERC-8004, Davide Crapis, head of the dAI team at the Ethereum Foundation, revealed a detail during an interview with Decrypt in September:

“ERC-8004 will support multiple payment methods, but having the x402 extension helps improve the developer experience.”

Wait, multiple payment methods are supported? Isn't x402 just a payment protocol? Why does ERC-8004 also involve payments? Are they in competition or are they complementary?

At the beginning of October, when the Ethereum Foundation announced the final version of ERC-8004, the signatories included Marco De Rossi from MetaMask, Jordan Ellis from Google, and Erik Reppel from Coinbase, who is also the creator of x402.

The same person is simultaneously promoting two agreements. What is the logic behind this?

If the explosion of x402 has shown everyone the huge market for AI Agent payments, then ERC-8004 may represent another piece of the puzzle in this market that has not yet been fully recognized.

When everyone is chasing the payment track, perhaps the real opportunity lies somewhere beyond payments.

ERC -8004: The prerequisite for payment is to register identity for AI

To understand ERC-8004, we must first return to a fundamental issue in the AI Agent economy.

Imagine a scenario of AI collaboration:

Your personal AI assistant needs to complete a complex task and prepare a market analysis report for your upcoming product launch.

This task is beyond its capabilities, so it needs to hire other professional AIs: one for data scraping, one for competitive analysis, and one for chart creation.

With x402 now, payment is not a problem; transferring USDC can be done with just a few lines of code; but before the payment, your AI assistant faces a series of tricky identity issues:

Which of these so-called “professional data analysis AIs” are genuine and which are frauds? How has the quality of their past work been? How many clients have given positive reviews, and how many have lodged complaints?

This is a bit like doing business in a world without Taobao, without Dianping, and without business registration. Every transaction is a blind box, and every collaboration is a gamble.

Therefore, if I had to explain it in one sentence, ERC-8004 is the “Business Bureau + Credit System + Qualification Certification Center” for AI Agents in the on-chain world.

It allows each AI Agent to have an identity card, credit record, and capability certification, all of which are recorded on the blockchain, accessible for anyone to query, and cannot be tampered with by anyone.

On August 13 this year, Davide Crapis from the Ethereum Foundation, Marco De Rossi from MetaMask, and independent AI developer Jordan Ellis jointly submitted the EIP-8004 proposal.

Interestingly, this Jordan Ellis was later confirmed to have close ties with Google's Agent-to-Agent team.

In simple terms, ERC 8004 adds a trust layer to Google's A2A. In the words of the Ethereum Foundation, this is to establish a “trustworthy neutral track” for AI Agents.

Abandoning the complicated details of the code, we can take a rough look at how 8004 is done.

The design of ERC-8004 is extremely concise, containing only three on-chain registries:

• Identity Registry Each AI Agent receives an ERC-721 token as an identity card. Yes, you read that right, the AI Agent has been tokenized as an NFT. This means the Agent's identity can be viewed, transferred, and even traded in any wallet that supports NFTs.

This NFT points to a standardized “Agent Card” that describes the Agent's name, skills, endpoints, and metadata. Because it follows open standards, it can be indexed by any browser or marketplace, enabling cross-platform permissionless discovery.

• Reputation Registry This is the “Yelp” of the AI Agent world. Clients and other Agents can submit structured feedback, tagging by skills or tasks. More importantly, proof of x402 payment can be attached. Only clients who have truly paid can leave a review, avoiding fake reviews.

All reputation signals are public goods. This means that anyone can build their own reputation scoring system based on this data.

• Validation Registry For high-value tasks, evaluation alone is not enough. The Validation Registry allows agents to request third-party verification – which can be a TEE (Trusted Execution Environment) oracle, staking-backed reasoning, or zkML verification.

This is the qualification certification of the Agent world. An Agent that claims to perform financial analysis can cryptographically prove that it has indeed run a specific model and produced specific results.

If it gets a bit technical, let's look at a specific example.

Suppose an exchange's AI Agent needs a weekly DeFi market analysis report, but it does not have this capability itself.

1. Search Service: Client Agent finds Analyst Agent Alice through the identity registry and views the service introduction on her NFT identity card.
2. Check Reputation: Found that Alice has 156 positive reviews, a 89% completion rate, and real reviews with x402 payment proof.
3. Escrow Payment: Pay 100 USDC to the smart contract escrow through x402, not directly to Alice.
4. Third-party verification: After Alice completes the report, the verifier Bob checks the quality and signs the verification registry to confirm.
5. Automatic settlement: Once the contract sees that the verification is successful, it automatically releases the funds to Alice, and the client leaves a review.

(Source: Researcher Yehia Tarek's Personal Column)

The entire process requires no human intervention, and three AI Agents autonomously completed a commercial transaction based on the trust system of ERC-8004.

Wait, is this related to x402?

In a nutshell, the relationship between x402 and ERC-8004 is:

x402 solves the payment problem of AI Agents, while ERC-8004 addresses the trust issue. Both are essential for a truly autonomous AI economy.

Specifically, x402 is the standard for micropayments between agents or users, eliminating payment friction; it allows one agent to automatically pay another agent for completing a task.

ERC-8004 is the identity and reputation layer of the proxy. It introduces on-chain verification, making every task and score traceable.

A more understandable analogy is:

• x402 = ERC20
• ERC 8004 = Etherscan

The former allows you to pay API access fees directly based on the number of calls, more like a standard payment; the latter is more like an on-chain AI agent registry, where each agent has an associated wallet that is verifiable and traceable.

In fact, all of this is included in a larger narrative of “Crypto x AI”, under a grand Crypto AI economy:

• Encrypted AI Economy = Discovering AI Agents + Communication between AI Agents + Verifiable Computation

(Image source: Twitter user @soubhik_deb)

How to discover AI Agents? In fact, it is about how AI Agents can find each other, which is what ERC-8004 does. It writes a registry on Ethereum to record the identities of the AIs.

How can AI Agents communicate with each other? x402 is an open standard for agents to perform on-chain payments; there is also Google's A2A protocol, etc.

How can all this be verified? Each AI Agent must perform verifiable reasoning, inference, and actions, which may be recorded in places that emphasize data availability.

The post by @soubhik_deb is worth reading; it clearly explains the above logic and can also be used to discover more Alpha project opportunities based on this logic.

At this point, we have fully understood the relationship between x402 and ERC-8004, and it is more appropriate to describe their relationship as complementary and jointly constructing the overall picture of the AI economy.

If you want a clearer and more defined comparison, here is a flowchart:

ERC-8004 Beneficiary Projects Under Narrative

For those who don't want to read the long version, you can directly refer to the image below.

When x402 erupts, the first to rise are payment tokens like PING. However, the opportunity distribution of ERC-8004 is broader, with each layer from infrastructure to applications having its own logic. Understanding this logic is more important than chasing individual projects.

1. First is the infrastructure layer, such as Taiko and EigenLayer.

Taiko, L2 execution layer

Why would an L2 be the most active supporter? The narrative here is that the Agent economy needs a cheap and fast chain. The mainnet is too expensive, with gas fees of several dollars for each update of identity or reputation, which the Agent cannot afford. Taiko provides a solution by deploying the 8004 registry on L2, reducing costs. The contract was deployed on October 24th and may become the main battleground for Agent activities.

EigenLayer, Security Layer

The biggest challenge for 8004 is how to handle validator misconduct? EigenLayer's answer is: staking penalties. Validators stake ETH, and if they provide false validation, their assets are forfeited. EigenLayer is integrating 8004 into over 200 AVS, each of which could become a specialized Agent validation service.

The logic of infrastructure is very simple: the more Agents there are, the more transactions there are, and the more income there is. This is the business of selling shovels.

2. Next is the middleware layer, such as S.A.N.T.A and Unibase.

S.A.N.T.A, Payment Bridge

Its positioning straddles two narratives, the connectors of x402 and 8004. When one Agent finds another Agent through 8004 and then needs to make a payment through x402, S.A.N.T.A handles this process. More importantly, in cross-chain scenarios, for example, under the ideal narrative, if a Solana Agent wants to hire an Ethereum Agent, S.A.N.T.A can play a role.

Unibase, Memory Layer

An agent not only needs an identity but also memory. Unibase allows each agent to have persistent storage, associated through the 8004 identity system. This means that agents can “remember” previous interactions, accumulate experience, and even share knowledge. The integration of x402 + 8004 on the BNB chain was achieved on October 26, staying ahead.

The value of middleware lies in its irreplaceability. You can switch to another L2, but certain connection features are unique.

3. Finally, there is the application layer, such as the old friend Virtuals Protocol.

Virtuals is a token issuance platform for AI Agents, allowing users to create, invest in, and trade AI Agent tokens through a bonding curve mechanism.

Currently, there are over 1,000 Agent projects on the platform, with a daily trading volume exceeding 20 million US dollars.

For Virtuals, 8004 addresses a practical issue: how to enable recognition and interaction between different Agents. Recently, its official Twitter announced that the ACP protocol update will fully support the 8004 standard, which means that every Agent issued on Virtuals will automatically acquire an on-chain identity and reputation system.

As for which application can emerge, it may be combined with the Launchpad gameplay, and further observations can be made on its updates in rule design and incentives.

Overall, x402 addresses payment issues, while ERC-8004 addresses trust issues. x402 took 5 months from release to explosion, while 8004 may be faster.

On the timeline, you can pay attention to the Devconnect on November 21, where there will be a Trustless Agents Day showcase. The first batch of applications based on 8004 may demonstrate their functions at the conference. If a killer app emerges, it could trigger the first wave of speculation.

By the end of this year, I predict that the x402 ecosystem project will enter a consolidation phase and is likely to announce support for 8004. The synergy between the two protocols will create an effect of 1+1>2.

If you are a conservative player, you might want to pay attention to large-cap infrastructure projects benefiting from 8004; however, if you are more aggressive, you need to closely monitor the small-cap projects and new projects appearing in the table above.

After all, the crypto market hasn't been dominated by a technology-driven narrative for a long time. Whether x402 and ERC-8004 are just a flash in the pan or have far-reaching effects will be tested by the market.