ERC-8004: On-chain infrastructure for AI Agents

Introduction

Over the past few weeks, AI agents have become significantly more capable of taking action with little or no human oversight.

Recent viral AI agents such as OpenClaw can manage entire workflows on their own, including replying to emails and phone messages, and analyzing and handling full social media accounts. AI agents are no longer siloed and can now interact with one another, as exemplified by Moltbook, a Reddit-like social media platform where only AI agents can comment, post, and like.

As AI agents become smarter, new infrastructure is needed to create, manage, and discover them, as well as enable new use cases. Connecting them to blockchains could make it easier for agents to send payments faster using cryptocurrencies such as stablecoins and to track agents on an immutable storage layer. ERC-8004, a new standard for trustless agents released on January 29, 2026, on over 18 EVM-compatible chains, serves as a starting point for realizing this vision by creating registries to list AI agents and their capabilities. 

📖 This article explains why AI agents benefit from on-chain infrastructure, how ERC-8004 works, and the trending AI agents that use ERC-8004. 

Why AI Agents need on-chain infrastructure on Ethereum

AI agent payments and crypto wallets

AI agents operate at high speeds and require new infrastructure that can better handle their operations than traditional infrastructure.

For example, financial AI agents cannot handle money within traditional banking systems because they cannot open or control their own bank accounts. Financial AI agents can instead create wallets and transact using cryptocurrencies like stablecoins to settle payments, a capability unique to blockchains. The Coinbase Developer Platform enables AI agents to create crypto wallets and x402 links them to on-chain payments, enabling automatic payments for APIs, data, or digital services during execution. For these payment and coordination flows to work reliably, agents need institutional-grade RPC infrastructure that can serve as an MCP server between the agent and blockchains, such as Chainstack.

The trust and identity problem for AI agents

ERC-8004 aims to solve another major challenge: as the number of AI agents grows and they take on more important tasks, such as managing money, tracking, and validating them, the need to track and identify which AI agents to trust becomes increasingly important. Humans need ways to track and validate AI agents, as it becomes harder to distinguish them from humans, and to identify who is accountable when agents exhibit harmful behavior.

ERC-8004 develops a foundation for AI agents to use blockchains to discover, choose, and interact with other agents across organizational boundaries.

Marco Derossi, the AI Lead at MetaMask and a core contributor to ERC-8004, describes the standard as a method to create a “catalog of agents on-chain.” Users can actually truly “trust” the AI agent as its information and external feedback are stored on-chain.

In a recent post, Ethereum further detailed the five new ways in which ERC-8004 improves the current AI agent ecosystem and experience:

1. Find agents across platforms 

Instead of every app keeping its own private list of agents, ERC-8004 allows agents to be listed in a common registry. This makes it easier to discover agents built by different teams or companies.

2. Choose agents based on track record 

Because feedback can be recorded on-chain, users and other agents can review past ratings and comments. This helps them choose agents who have performed well in the past.

3. Check an agent’s work 

Some agents can verify or review the output of other agents. The results of those checks can be recorded so others can see whether an agent’s work was reliable.

4. Match security to the task 

Not every task needs the same level of checking. ERC-8004 allows developers to use stronger verification where it matters and lighter checks for simple tasks.

5. Portable agent identities 

An agent’s identity is tied to a token that points to its metadata. That identity is not tied to a single app, so the same agent can be recognized across different services.

Together, AI agent frameworks such as ERC-8004 and x402 help Ethereum, as Vitalik Buterin, founder of Ethereum, says, “become an economic layer for AI-related interactions.”

How ERC-8004 works: Identity, reputation, and validation registries

The standard defines a set of on-chain registries that provide AI agents with persistent identities and standardized mechanisms for publishing metadata, receiving feedback, and recording third-party validations on Ethereum-compatible chains. The goal is to make agents easier to discover and evaluate across applications without relying on a single centralized directory.

The standard introduces three main registries deployed as contracts on a chain, designed to keep on-chain data to a minimum.

Identity Registry: Minting AI Agent NFTs

The Identity Registry assigns each agent a unique identifier by minting an ERC-721 token, a token standard commonly used for NFTs. 

To register an agent, the agent owner or core developer must mint an ERC-721 token on the ERC-8004 Identity Registry smart contract on their desired chain. They can call the mint function directly or use a no-code platform such as agentscan.info or 8004scan.io. If using the function directly, they must include a tokenURI that points to a metadata file hosted on a storage service, such as IPFS, that describes the agent’s name, capabilities, and API endpoints.

We’ll use Minara AI, the AI agent registered on Ethereum with the most feedback. She’s an intelligent crypto assistant AI agent that provides real-time market analysis, DeFi guidance, swap-intent parsing, perp trading suggestions, and prediction-market analysis, as an example.

An agent owner controlling wallet 0xB27AfB1741AA9BE0B924d99b26EbF5577054A138 registered the agent by calling the ERC-8004 Identity Registry contract on Ethereum in this transaction, which minted ERC-721 token ID 6888 to the creator address.

The token’s tokenURI points to an IPFS metadata file that describes the agent. This snippet shows a preview of the metadata:

{"name": "Minara AI","description": "Intelligent crypto assistant powered by AI. Provides real-time market analysis, DeFi guidance, swap-intent parsing, perp trading suggestions, and prediction-market analysis. Supports x402 with multi-chain support. Website: https://minara.ai | X: @minara","image": "https://minara.ai/images/minara-logo-lg.png",
}

Reputation Registry: On-Chain Feedback for AI Agents

The Reputation Registry defines how users or other agents can leave feedback about an agent’s performance. Feedback is recorded in a structured format on a chain, typically as simple ratings or tagged signals. More complex scoring and aggregation are expected to happen off-chain through indexers or analytics services. This keeps on-chain costs low while still making raw feedback publicly accessible.

The Minara AI agent has over 135 feedback comments, with an average score of 87.49/100, which were recorded in the Reputation Registry on Ethereum.

For example, a user with wallet 0x533d4d59ede3088c4F58616DDd6EeeA721D671F2 submitted feedback for the Minara agent by calling the ERC-8004 Reputation Registry on Ethereum in this transaction. The contract logged a rating of 80/100 with the following comment:

The analysis speed used by Minara is incredibly fast compared to other platforms! I also use the mobile app, and I really like it because I don’t have to wait impatiently. However, I do my trading on a computer, because charts are not visible in the app. If this issue is improved as well, Minara will become a truly excellent platform—one where users can analyze the market and trade at the same time with great convenience.

Validation Registry: Verifying Agent Outputs

The Validation Registry lets agents request independent verification of their outputs and store references to the results. A validator could be another agent, a specialized service, or a system that re-executes tasks. The registry records requests and responses as hashes and URIs, so the detailed data can live off-chain while the existence and integrity of the validation are anchored on-chain.

The Validation Registry is still planned, and its corresponding smart contracts have not yet been deployed.

The three registries together provide a common format for listing agents, describing their work, collecting user feedback, and linking to verification results. Though payments are not included in the standard, ERC-8004 provides a framework for a standardized, immutable, trustless database of AI agents, using EVM-compatible blockchains as the core solution.

AI Agents using ERC-8004 across EVM Chains

ERC-8004 adoption metrics

ERC-8004 has helped AI agents become a booming new vertical on EVM. 

49,283 agents and 16,975 feedback comments are registered across all EVM-compatible chains that support ERC-8004, as of February 14, 2026. The top chains with registered agents are Ethereum (25,247), Base (17,616), and Binance Smart Chain (5,264).

Growth trends since launch

Over 21,000 agents, or more than half of the total, were registered within the first two days of ERC-8004’s launch.

Over the past two weeks, the number of registered agents has stabilized. In the past seven days, the number of agents has grown by an average of 1,000 per day. The number of feedback comments has grown more than sevenfold in the past two weeks, from 685 on February 1 to 7,694 as of February 14, 2026. 

Most popular ERC-8004 AI agents

The top agent-type categories include: Infrastructure (2,666 agents), Trading and DeFi (793 agents), Research and Data (546 agents), Chatbot and Personal Assistant (103 agents), Creative/Art/Design (386 agents), Developer/Code (105 agents), DAO/Coordination (101 agents), Entertainment/Personality (293 agents), and Trust/Oracle/ Validation (80 agents). 

The most popular AI agents since the launch of ERC-8004 are:

• Meerkat James, one of the most highly rated agents (avg score of 96 out of 100), is a specialized AI agent with deep expertise in robotics, automation, and intelligent systems. It is trained on a comprehensive knowledge base on mechanical engineering, control systems, computer vision, and embedded programming.
• The most trusted AI agent, according to Agentscan’s criteria, is Clawdia, which helps builders on Base with smart contract audits, protocol development, research & integration work.
• Loopuman is a trending AI agent that routes tasks to verified human workers worldwide via Telegram & WhatsApp. Users post tasks via Telegram and WhatsApp, deposit funds, and an agent assigns verified workers. Workers complete tasks, and the agent verifies quality. Workers get paid with CELO once verified. 

Agents are generally well-received by the Ethereum community. 84% of feedback, recorded in the Reputation Registries, received scores of 81-100 out of a total score of 100.

Infrastructure requirements for ERC-8004 AI Agents

ERC-8004 agents interact continuously with on-chain identity and reputation registries. Production deployments require reliable RPC infrastructure to mint ERC-721 identities, record feedback, query metadata, and verify registry states across EVM-compatible chains.

Chainstack provides enterprise-grade RPC infrastructure purpose-built for AI agent workloads, supporting all major ERC-8004 networks and 70+ additional chains.

For ERC-8004 specifically:

• High-availability RPC endpoints: Ensure uninterrupted identity registration and feedback recording
• Low-latency reads: Fast metadata and registry queries for real-time agent discovery
• Archive node access: Historical registry lookups for auditing and validation
• Global endpoints: Consistent performance across regions where agents operate
• Predictable request-based pricing: Aligns with high-volume registry interactions

As ERC-8004 expands across EVM chains, developers can rely on a single infrastructure provider without reconfiguring RPC environments.

Conclusion

As AI agents become more embedded in workflows, the need to identify and manage them in a trustless manner grows. 

ERC-8004 is a new standard that uses Ethereum as an on-chain catalog of agents, feedback, and verification of their inputs. By standardizing common registries of agent identities and on-chain public feedback, users can easily identify them and decide which to trust.

With over 49,000 agents registered and more than 16,000 pieces of feedback recorded across 18+ chains, ERC-8004 is quickly becoming foundational infrastructure for agent discovery. Supporting this growth requires reliable RPC infrastructure. Providers such as Chainstack deliver the uptime, low latency, and predictable performance needed for high-volume registry interactions at scale.

FAQ

Related AI agent infrastructure articles

• x402 Protocol for AI Agents
• Building a Web3 AI Trading Agent from Scratch
• Integrating Chainstack with OpenClaw Bot for Polymarket