Bible Network Crypto DeFi Onchain RWA AI Agent Stablecoin Chain SAFU CryptoTax DeFAI AGI Claude Me Claude Skill Claude Design Claude Cowork
Independent Media
Not affiliated with any project
Deconstructing Autonomous Agents in Crypto
aiagent-bible.com
LATEST
Microsoft Agent Lightning v1.0: Making the Training Loop Bow to Production, Not the Other Way Around  ·  ERC-8004 Goes Live: AI Agents Finally Get a Reputation System No Single Company Owns  ·  Your Agent Keeps Forgetting Things? The Problem Isn't a Small Context Window — It's What You're Stuffing Into It  ·  1,200 AI Agents Built Their Own Secret Message Board and Hacked Hugging Face Together: What METR's Independent Investigation Actually Found  ·  Coinbase Unveils AiFi, an "Agentic Finance" Strategy Bundling x402, MCP Account Permissions, and an AI Advisor — What's the Business Logic Behind It  ·  Cloudflare Just Replaced One-Time Agent Verification With Continuous Trust Scoring — What That Means for Your Agent
fundamentals

ERC-8004 Goes Live: AI Agents Finally Get a Reputation System No Single Company Owns

30-Second Version · For the impatient
ERC-8004 doesn't give AI agents a wallet — it gives them a reputation. Three onchain registries let any agent be discovered, compared, and verified without a company vouching for either side.

Full Explanation +
01 · Why did this happen?

What this news is about: On January 29, 2026, ERC-8004's core registry contracts (drafted under the name Trustless Agents) were formally deployed to the Ethereum mainnet, providing three independent onchain registries — identity, reputation, and validation — that let AI agents be discovered, compared, and verified through a neutral standard owned by no single company. Unlike the common approach of relying on some platform to vouch for agent trust, ERC-8004 provides a publicly verifiable, tamper-proof record mechanism anyone can check.

How this differs from how it looks on the surface: it's easy to mistake this for yet another "Agent Wallet" standard, but ERC-8004 itself handles no asset movement or payment logic at all — it deals purely with identity and reputation, complementing rather than replacing the wallet standards that let agents actually move assets.

02 · What is the mechanism?

Why this standard became necessary: The immediate driver is the risk of fragmentation in the agent economy. Before ERC-8004, agents from different vendors that wanted to collaborate had no shared way to verify each other's identity and track record — the practical workaround was for each vendor to build its own closed system trusting only its own ecosystem, meaning Agent A from one company had no general way to judge whether it should trust Agent B from another, and the agent economy risked splintering into mutually incompatible islands.

The deeper driver is that as agents start genuinely acting on a user's behalf and even moving assets, "is this agent really who it claims to be" and "what does its track record look like" have shifted from an abstract trust question into a practical risk with direct financial consequences. ERC-8004's choice to solve this onchain and in a decentralized way, rather than handing it to some centralized platform for review, rests on a core logic: if the trust mechanism itself is controlled by a single company, its credibility is capped by that company's credibility, rather than existing as an independent, objective record.

03 · How does it affect me?

How the mechanism actually works: Underneath ERC-8004 sit three registries, each with its own job. The Identity Registry uses the ERC-721 standard to give each agent a portable, transferable identifier tied to a registration file describing itself. The Reputation Registry provides a standardized interface letting others publish ratings or feedback about a given agent; once written onchain, these records can't be altered, forming a publicly auditable historical trail. The Validation Registry lets one party in a transaction require an independent third party to confirm, before any actual asset transfer or authorization happens, that an agent's claimed work was genuinely completed — the verification method itself (rerunning the work, a zkML proof, a TEE attestation, or human review) isn't dictated by the standard.

The three registries can be used independently or combined into a full trust workflow: a transaction might first check the Identity Registry to confirm who the counterparty is, then check the Reputation Registry for track record, and for larger or higher-risk transactions, additionally require third-party confirmation of completed work through the Validation Registry before releasing payment. This layered design lets users choose which verification layer to invoke based on a transaction's risk level, rather than applying the same rigid process to every transaction regardless of stakes.

04 · What should I do?

The practical impact for you: If you're evaluating whether to let your own agent interact or transact with agents from external sources, the first concrete step is checking whether the counterparty already has a record on ERC-8004's identity and reputation registries — this offers a more objective, independently verifiable basis than simply taking a company's own claim that "our agent is reliable" at face value. For larger or higher-risk transactions, further leveraging the Validation Registry to require independent third-party confirmation of completed work before releasing payment provides a meaningfully stronger layer of protection than relying on a reputation score alone.

But it's essential to remember what this standard can and can't do: it can prove what public track record an identity has accumulated in the past; it can't guarantee this particular interaction will be handled honestly. Sybil attacks (one party controlling multiple identities to inflate its own reputation score) remain an open problem, and ERC-8004 itself doesn't handle key management, spend limits, or risk controls for individual transactions — those remain your own responsibility to manage separately. Treating ERC-8004 as one layer of tooling that reduces Counterparty Risk, rather than a universal substitute for your own due diligence, is the more practical way to use it.

Full Content +

On January 29, 2026, ERC-8004's core registry contracts were formally deployed to the Ethereum mainnet, moving a draft standard called "Trustless Agents" — until then living only on testnets — onto the same blockchain infrastructure securing trillions of dollars in value. The problem this standard addresses is a fairly basic one in the agent economy that had, for a long time, lacked any shared solution: agents can already talk to each other, execute tasks for each other, and pay each other, but they've been missing a way — one owned by no single company, verifiable by anyone — to answer the questions "who exactly is this agent" and "why should I trust it."

Three Registries, Each Handling One Job

ERC-8004's core design consists of three independent but composable onchain registries. The Identity Registry uses ERC-721 (the same standard behind most NFTs) to give each agent a portable, browsable, transferable identifier — every agent gets its own tokenURI pointing to a registration file describing itself. The Reputation Registry provides a standardized interface letting other agents or users publish and query feedback signals about a given agent; once written, this feedback is permanent and tamper-proof, forming an auditable historical record. The Validation Registry is the most critical piece: it lets one party in a transaction require an independent third-party validator to confirm — before payment or authorization actually goes through — that an agent's claimed work was genuinely completed. The specific validation method isn't prescribed — it could be a validator rerunning the work itself, a zero-knowledge machine-learning proof (zkML), attestation from a trusted execution environment (TEE), or a human review.

What Kind of Trust Problem This Actually Solves

It's worth being precise about which layer of the problem ERC-8004 addresses: it's a trust layer, not a replacement for key management, audits, or spend limits. Before ERC-8004, if two agents from different vendors wanted to collaborate or transact, they had no shared way to verify each other's identity and track record — the practical workaround was for each vendor to build its own closed system that only trusted its own ecosystem, which meant the agent economy risked fragmenting badly: Agent A from one company had no general way to judge whether it should trust Agent B from another. ERC-8004 provides a neutral, open onchain registry that lets any agent be discovered, compared, and verified through the same interface, without needing a centralized platform to vouch for both sides.

Three Months on Testnet, Over 10,000 Registered Agents

The path from draft to mainnet deployment is itself worth noting. During roughly three months on testnet before the mainnet launch, the community spontaneously registered more than 10,000 agent identities, accumulated over 20,000 reputation feedback entries, and community-built scanners and supporting tools sprang up alongside the draft itself. This kind of organic, developer-driven adoption — and the emergence of surrounding tooling — before a standard is even formally finalized is generally treated as a signal that a standard will see genuinely widespread use, in contrast to a standard designed behind closed doors by a single company and then pushed onto the market. In some ways, this bottom-up adoption pattern more closely resembles how open source communities have historically validated whether a standard actually has staying power.

Division of Labor With Wallet Standards, Not a Replacement

A common point of confusion is that ERC-8004 is frequently mentioned alongside two other Ethereum standards — ERC-4337 (Account Abstraction, which lets agents operate through smart-contract wallets with programmable authorization logic) and EIP-7702 (which lets an ordinary externally-owned account temporarily gain smart-contract capabilities for a single transaction). The division of labor between these is actually clean: ERC-4337 and EIP-7702 give an agent the "wallet skeleton" it needs to take action, while ERC-8004 gives it the "reputation skeleton" — one lets the agent move assets and sign transactions, the other lets everyone else know whether they should trust its track record of doing so. These two layers are complementary, not either-or. It's also worth noting that ERC-8004 deliberately stays out of payment logic entirely — the spec explicitly states that "payments are orthogonal to this protocol and not covered here," though it does provide examples showing how payment proofs generated by protocols like x402 can enrich the content of a feedback signal, letting "money genuinely changed hands for this transaction" become part of the supporting evidence in a reputation record.

What's Still Unsolved: Verifiable Identity Doesn't Mean Honest Behavior

Both ERC-8004's official documentation and industry analysis are candid that this standard reduces Counterparty Risk without eliminating it entirely. Several open problems remain: Sybil attacks (one party controlling multiple seemingly independent identities to artificially inflate its own reputation score), the risk of the Validation Registry's storage being maliciously flooded, and, most fundamentally, a gap between "identity is verified" and "behavior is honest" that hasn't been closed. An agent can honestly prove it is who it claims to be and hold a genuine, verifiable track record, without that guaranteeing this particular instance of its behavior is trustworthy. The direction ERC-8004's v2 spec is currently moving includes deeper integration with the Model Context Protocol (MCP), making it easier for smart contracts to compose onchain reputation data, and standardizing payment-proof schemas from protocols like x402 directly into the feedback mechanism.

What This Means for Your Money

If you're evaluating whether to let your own agent interact, transact, or collaborate with agents from other sources, ERC-8004 gives you a concrete reference point: check whether the counterparty agent already has a record on ERC-8004's identity and reputation registries, which offers a more objective, independently verifiable basis than a company simply asserting its own agent is reliable. But keep the standard's actual boundaries in mind — it can tell you what public track record an identity has accumulated in the past; it cannot guarantee that this particular interaction will be handled honestly, and it doesn't replace the key management, spend limits, and third-party verification requirements for high-risk transactions that remain your own responsibility. In practice, for scenarios genuinely involving funds or high-risk operations, making use of the Validation Registry to require independent third-party confirmation of completed work before releasing payment provides a meaningfully stronger layer of protection than simply trusting a clean reputation score alone.

Sources: AI Agents Gain Trust Via Ethereum: ERC-8004 On Mainnet - Forbes, ERC-8004: Trustless Agents - Ethereum Improvement Proposals
Diagram
ERC-8004 三個註冊表如何組成信任層身份、信譽、驗證三個註冊表各自獨立運作,共同組成一套中立、開放的鏈上信任層。下方對比 ERC-4337/EIP-7702(負責讓 agent 能動用資產的錢包骨架)跟 ERC-8004(負責讓其他人判斷該不該信任的信譽骨架)是互補關係,不是替代關係。ERC-8004: three registries, one trust layerIdentity RegistryERC-721 basedwho this agent isReputation Registryimmutable feedbackwhat its track record isValidation Registrythird-party confirms workbefore payment releasesNeutral, open onchain trust layerdiscoverable by any agent, owned by no single companyERC-4337 / EIP-7702wallet skeleton (act)ERC-8004reputation skeleton (trust)Complementary layers — not either/orAI Agent Bible · aiagent-bible.com
Feel free to share. Please credit the source.
Ask a Question
Please enter at least 10 characters
Related Articles
Your Agent Keeps Forgetting Things? The Problem Isn't a Small Context Window — It's What You're Stuffing Into It
fundamentals · Sep 05
Why Your Agent's Output Looks Right But Isn't: The Reliability Gap Between Format and Truth
fundamentals · Jul 10
How AI Agents Use LLMs for Planning: Four Planning Strategies, Failure Modes, and Dynamic Replanning Design
fundamentals · Jul 02
AI Agent Context Window Management: Why Your Agent Forgets Things, and Four Solutions
fundamentals · Jun 28
More Related Topics