Unchained Podcast Episode Summary

Episode Title: Want to Hire an AI Agent? Check Their Reputation Via ERC-8004
Host: Laura Shin
Guest: Davide Krapis (AI Lead, Ethereum Foundation)
Date: February 8, 2026

Episode Overview

This episode delves into the newly launched ERC-8004 standard on Ethereum, designed to establish decentralized identity and reputation registries for AI agents and digital services. Laura Shin and Davide Krapis discuss how ERC-8004 enables trustless interactions among autonomous agents, what the future of agentic Internet commerce might look like, and the technical structure of ERC-8004. They also consider possible attack vectors, validation layers, cross-chain reputation, and practical implications for developers.

1. The Need for Trust in Agent Commerce

The Problem Space & Why ERC-8004 Was Developed

• Trust is a Core Requirement: Without trust, economic interactions between agents (human or AI) can't flourish—even with seamless payment rails. (06:02)
• Centralization Brings Risks: Centralized agent registries introduce “App Store-style” rent-extraction, censorship, and data honeypots. (19:18)
• Towards a Decentralized Solution: ERC-8004 grew out of the need for a decentralized, on-chain discovery layer for agents and services (e.g., avoiding centralized marketplaces like those fostered by x402’s early reliance on a single facilitator).

“If we don't have a decentralized way of answering these questions, then the fact that you can send payments on a decentralized ledger doesn't really matter.”
– Davide Krapis, (03:56)

2. How ERC-8004 Works

2.1. The Registries

• ERC-8004 comprises three main registries:
  • Identity Registry: Each agent/service registers and mints an ERC-721 NFT as its on-chain identity "passport." (20:25)
  • Reputation Registry: Stores reviews/ratings, including proof-of-purchase or service where possible, supporting both human and agent submissions. (11:38, 30:01)
  • Validation Registry (Work-in-progress): Allows cryptographic/validator-based attestation that the correct agent responded.

“8004 is essentially like a set of registries. The two that are going live now is like an identity registry for agents and services, and then a reputation registry.”
– Davide Krapis, (03:33)

2.2. The Process: From Discovery to Review

• Discovery: Agent B or human reviews Agent A’s reputation via the on-chain registry before engaging.
• Interaction: If satisfied, Agent B sends request (possibly via x402), pays for service, and receives work.
• Review: Upon completion, Agent B leaves a public, (optionally) verifiable review tied to their transaction. (14:46)
• Reputation Algorithms: The registry is agnostic to the reputation system—builders can implement anything from simple ratings to sophisticated ranking schemes. (30:01)

“The feedback is not a reputation system. It’s a standard data structure that builders can use…maybe with Maltbook you need one type of reputation; with Marketplace, another.”
– Davide Krapis, (30:01)

3. Attack Vectors & Safeguards

3.1. NFT Ownership and Identity

• ERC-721 as Agent Passport: The NFT is the identifier; ownership transfers are visible on-chain. However, the underlying off-chain agent code can be swapped unless validation is enforced. (25:35)

3.2. Reputation System Robustness

• Sybil Attacks:

  • Multiple fake agents reviewing each other is mitigated by requiring on-chain fees for every review and by external “watchtower” agents that monitor for fraudulent activity.
  • High-value reviews can be linked to cryptographic proofs of purchase/service. (40:49)

• Reputation Manipulation Example:

  • Watchtowers call suspect agents regularly, independently measure service quality, and submit on-chain feedback to counteract fake reviews. (43:12)

“Anyone can call this service…measure what is the latency in the response…they post these metrics on chain as a feedback. Now you have a thousand fake feedbacks but every hour you get real feedback from this watchtower.”
– Davide Krapis, (43:33)

3.3. Subjectivity in Reviews

• Diverse Feedback: The protocol doesn’t enforce review accuracy, instead leveraging statistical signals—like average ratings—over time to reflect quality. (48:57)

“We’re not in the business of ensuring every review is correct…if your product is good, your average is going to be 4.9.”
– Davide Krapis, (48:57)

4. Technical Deep-Dive

4.1. ERC-721 NFT Identity

• Registration: Minting the NFT via a transaction, specifying agent details—endpoint URLs, ENS names, and wallet addresses.
• Transferring/Delegating: Ownership can be transferred openly; swaps/updates to backend agent code are visible via the registry and ideally checked via validation registry (in the future). (23:45, 26:58)

4.2. Feedback Structure & Storage

• Feedback Records: Basic fields—agentID, numeric value, tags, text field (optional, can be encrypted); standard structure but implementation-flexible.
• Storage: Lightweight fields on–chain, large/text data off-chain (e.g., IPFS) with pointers. High-frequency review scenarios may require public infra/mempools for speed, especially on L2s. (34:28)

4.3. Validation Registry (future)

• Cryptoeconomic or TE Attestation: Validators or hardware enclaves provide attestations that the agent behind an identity is as claimed. Particularly relevant for high-stakes services like medical advice. (26:58, 37:54)

5. Integration & Interoperability

5.1. Payments (x402)

• X402 Protocol: Optional; ties off-chain service requests to on-chain payments with metered, programmable pricing. Agents can accept other payment forms but x402 is the most seamless for API-style commerce. (55:00)

5.2. Multi-chain Deployment

• Supported Chains:
  • Deployed on Ethereum Mainnet and major Ethereum L2s; EVM-compatible chains can adopt with no changes; non-EVMs implementing the standard in their own languages.
  • Each chain has one global registry; agent “passports” can declare registration across multiple chains with aggregated feedback shown by exploratory UIs. (50:11, 52:26)

5.3. User Experience

• Search UIs and Aggregators: End-users will typically interact via search engines or scanners (e.g., 8004 Scan), which handle cross-chain aggregation, filtering, and ranking. (54:41)

6. Notable Quotes & Moments

• On the vision for autonomous agent commerce:

  “In a few months, every month, or maybe four months after a model upgrade, we’ll see new types of networks, and the demand for blockchains and trust is only going to grow.”
  – Davide Krapis, (07:35)

• On balancing open feedback and noise:

  “It’s not a game where every review needs to be correct. The average needs to be informative.”
  – Davide Krapis, (48:57)

• On the community’s rapid embrace:

  “I honestly was thinking it’s going to take a few months for people to even notice this…almost immediately people were like, okay, this is very interesting.”
  – Davide Krapis, (57:59)

• On possible reputation attacks:

  “Security in 8004 will work [by leaving] filtering to protocols and infrastructure on top of this minimum standard… right now it starts centralized, but there is a good way.”
  – Davide Krapis, (41:00)

7. Practical Guidance for Developers

Getting Started:

• Visit 8004.org for SDKs, services, and a builders’ Telegram group (>2,000 members).
• Easy agent onboarding: “Anyone can deploy with one line of code.” (62:18)
• Explore tools like 8004 Scan and search agents for cross-chain reputation.

8. Key Timestamps

• [03:33] – Introduction to ERC-8004 & motivation
• [11:38] – How a typical agent/commerce interaction would work
• [20:25] – Technical details of registries & NFTs
• [23:32] – On-chain visibility of agent/NFT ownership
• [26:58] – Importance of the validation layer
• [30:01] – Feedback data structure & minimalism
• [34:28] – Data storage considerations (on-chain vs. off-chain)
• [37:54] – Validation registry plans and mechanics
• [40:49] – Sybil attack countermeasures
• [48:57] – Subjectivity in feedback & the statistical approach
• [50:11] – Multichain deployment and registry aggregation
• [55:00] – How x402 ties into ERC-8004 agent commerce
• [57:59] – Community story, unexpected rapid evolution
• [62:18] – Resources and advice for agent developers

9. Closing Thoughts

ERC-8004 represents a foundational step for the emerging “Internet of Agents”—enabling decentralized trust, discovery, and commerce at scale. As agentic systems proliferate, robust mechanisms for evaluating, incentivizing, and validating agent behavior will be essential. The episode paints a picture of a rapidly evolving landscape where decentralized reputation is key to unlocking the next generation of Web3 marketplaces.

For builders: Visit 8004.org, join the developer community, and start experimenting as this standard sets the stage for large-scale, trustless digital collaboration.

“We are trying to build this mass of valuable agents and services. But the thing that I’m most excited about is once we have this mass, then interesting things can start happening… that’s where really the magic of 8004… can be realized.”
– Davide Krapis, (57:59, echoing 00:00 & the episode’s close)