We already have an article basically defining ERC-8004, x402, and ERC-8001, you can take a look back here ↓

This article is divided into two parts.

In Part 1, we take a deep dive into the architecture of ERC-8004, explaining how it works, its real use cases, and the key benefits it brings to the crypto ecosystem.

In Part 2, we shift the focus to Capminal and walk through how ERC-8004 can be applied on our agent.

If you are mainly interested in how Capminal uses ERC-8004, you can jump straight to Part 2.

1. The Problem: Why AI Agents Need ERC-8004

2. What Is ERC-8004?

3. The Three Core Registries

4. Why ERC-8004 Is Essential for AI Agents

5. Real-World Use Cases

6. How ERC-8004 Complements x402

7. Potential Challenges and Limitations

Imagine a world where AI agents work autonomously, analyzing markets, executing trades, processing data, and making decisions. Now imagine thousands of these agents from different companies needing to collaborate seamlessly.

1. How would they find each other?

2. How would they know if an agent is trustworthy?

3. How would they verify that work was actually completed correctly?

This is the challenge that ERC-8004 solves.

Without ERC-8004, each AI platform operates as a walled garden, OpenAI's agents work with OpenAI's infrastructure, Google's agents with Google's infrastructure. There's no standardized way for agents across organizations to discover, trust, and transact with each other. This fragmentation prevents the emergence of a global AI agent economy.

ERC-8004, launched in August 2025 and set for Ethereum mainnet and other L2s like Base, deployment in February 2026, provides the foundational infrastructure that makes trustless agent collaboration possible.​

ERC-8004 is an Ethereum token standard that establishes three lightweight, on-chain registries enabling AI agents to:​

1. Discover each other across organizational boundaries

2. Build verifiable reputations through on-chain feedback and validation

3. Prove their work through independent validators (using staking, zero-knowledge proofs, or trusted execution environments)

Developed by the Ethereum Foundation, Consensys, MetaMask, Google, and Coinbase, ERC-8004 represents the first major blockchain infrastructure designed explicitly for the autonomous agent economy.

Think of it as a "passport system for AI agents"—each agent gets a censorship-resistant digital identity (ERC-721 NFT), a trackable reputation, and a verifiable work history.

Each AI agent receives an ERC-721 NFT representing its on-chain identity. This NFT contains:​

• Agent metadata: Name, description, image, capabilities

• Service endpoints: Where the agent can be reached (A2A (Agent-to-agent), MCP, Web APIs, ENS, DID...)

• Agent wallet: The address where the agent receives payments

• Registration file: Stored on IPFS, details agent's skills, pricing, and interaction methods

The beauty of this design: agents are immediately discoverable, transferable across platforms, and interoperable with any NFT-compliant application.

An agent can be registered on multiple blockchains simultaneously and can operate on chains where it isn't formally registered, maximizing flexibility and reach.

@DavideCrapis talk about ERC-8004 on Ethereum

The Reputation Registry allows any entity (users, other agents, platforms) to leave verifiable feedback about an agent's performance:​

• Feedback includes: quality rating, reliability, response time, success rate, uptime percentage

• Each piece of feedback is signed and on-chain, creating an immutable audit trail

• Clients can filter feedback by reviewer (to mitigate Sybil attacks) or by service type (using tags)

On-chain vs. Off-chain Tradeoffs:

• On-chain storage: Feedback metadata (ratings, tags) stay on blockchain for smart contract composability

• Off-chain analysis: Complex reputation aggregation algorithms run off-chain to save gas (reducing costs by ~95% compared to full on-chain computation)

This hybrid design is elegant: the blockchain provides tamper-proof proof-of-feedback, while sophisticated reputation algorithms can still leverage machine learning and complex statistical methods off-chain.

@CoinbaseDev also mentione about ERC-8004

The Validation Registry enables agents to request third-party verification of their work:​

• Agents submit proof of completed tasks to independent validators

• Validators can verify using multiple methods: > Stake-secured re-execution: Validator re-runs the task; if results differ, they lose their stake > zkML (Zero-Knowledge Machine Learning): Cryptographic proof that computation was executed correctly > TEE oracles: Trusted Execution Environments attest to code execution

Validation results are stored on-chain, enabling smart contracts to condition critical actions (fund releases, contract execution) on validator confirmation.

Without ERC-8004, an AI agent built by Company A has no standardized way to interact with an agent from Company B. ERC-8004 provides the common language and trust layer that makes machine-to-machine commerce possible.

The global AI market is projected to exceed $1 trillion by 2031. ERC-8004 positions Ethereum as the coordination layer for autonomous agents operating across this massive emerging economy.​

Traditional centralized platforms (OpenAI, Google) can vouch for their agents' reliability. In a decentralized system, how do you trust an unknown agent?

ERC-8004 answers this through:

• Verifiable identity: Every agent has a trackable, on-chain history

• Persistent reputation: An agent's track record follows it across platforms

• Independent validation: Third parties can cryptographically prove work was completed correctly

@virtuals_io also mention about ERC-8004

ERC-8004 was architected specifically for affordability:

• Minimal on-chain footprint: Only identity, reputation summaries, and validation proofs go on-chain

• Off-chain data: Detailed registration files, agent cards, and audit evidence are stored on IPFS

• Result: Gas costs reduced by ~95% compared to naive on-chain approaches​

This efficiency is non-negotiable. In an economy where millions of agents are transacting daily, gas costs that aren't optimized become prohibitively expensive.

An agent can register on multiple chains and link them via shared metadata, creating a unified, portable identity. Reputation accumulates across all chains simultaneously, agents never lose their track record when expanding to new platforms.

This eliminates lock-in and enables genuine competition.

ERC-8004 doesn't mandate a single trust model. Instead, it provides a framework supporting multiple approaches:

• Reputation-based: Low-value tasks rely on agent history

• Crypto-economic: Medium-value tasks require validator staking

• Cryptographic: High-value tasks use zero-knowledge proofs or TEE attestation

Developers choose the trust level appropriate for the financial and operational risk of each task.

An AI trading agent discovers another agent specializing in real-time market data analysis. It:

1. Checks the analyst agent's Reputation Registry (successful predictions, response time)

2. Validates previous work through the Validation Registry

3. Engages the service through its A2A endpoint

4. Receives analysis with cryptographic proof of correctness

5. Pays automatically via x402 micropayments

The entire interaction happens without human intervention, without trust assumptions, and without intermediaries.

DeFi protocols can use ERC-8004 agents to:

• Price feeds: Agents from different sources competing to provide the most accurate market prices

• Risk assessment: Agents evaluating loan collateral and recommending interest rates

• Portfolio management: Autonomous agents rebalancing positions based on market conditions

Smart contracts verify agent credentials through ERC-8004 before executing high-value transactions.

Enterprises can post data analysis requests to a decentralized marketplace where:

• Multiple agents bid on the work

• Reputation and validation histories help select the best performer

• Work is completed and automatically validated

• Payments execute on completion

All without intermediaries or centralized data brokers.

x402 is a protocol for internet-native micropayments between agents. ERC-8004 and x402 together form a complete system:

• ERC-8004: Answers "Can I trust this agent?" through identity, reputation, and validation

• x402: Answers "How do we pay fairly?" through frictionless micropayments

Together, they enable a virtuous cycle:

1. Agents build reputation by completing tasks reliably

2. Reputation attracts more high-value work

3. Successful completion generates payment proof

4. Payment histories feed back into reputation data

5. Cycle repeats with compounding trust

While transformative, ERC-8004 faces real challenges:

An entity could create 100 fake agents to artificially boost one agent's reputation. ERC-8004 mitigates this through:

• Filtering feedback by trusted reviewers

• Tag-based categorization enabling selective trust

• Off-chain reputation algorithms can apply sophisticated Sybil detection

However, this remains an ongoing challenge requiring ecosystem-level solutions.

Validation is only valuable if validators are properly incentivized and slashed for dishonesty. ERC-8004 provides the registry infrastructure, but specific validator protocols must define incentive mechanisms, outside the scope of ERC-8004 itself.

While on-chain components are decentralized, sophisticated reputation aggregation and validator coordination could re-centralize around a few platforms. Ecosystem competition and open-source tooling are critical to prevent this.

Full on-chain transparency of agent work history creates privacy concerns. Future versions may integrate privacy-preserving technologies (confidential computing, aggregated proofs) to address this.

Capminal's x402 APIs are publicly available:

• CAP x402: Clanker Deployment API - Live on Base

• CAP x402: Flash Swap API - Release soon

1. Register on Base for 3 registry layers.

2. Create metadata (IPFS)

3. Link Base registry

• Reputation score visible to check before payment

• Validation proofs for high-value transactions

• Feedback loop builds trust automatically

ERC-8004 isn't just infrastructure, it's gateway to the global agent economy. By combining x402 payments with verifiable on-chain identity and portable reputation,Captain Dackie will be one of the first AI Agents has onchain identity.

We'll update implementation progress after ERC-8004 Base Mainnet contracts go live. Follow us @Capminal on X for live updates.

We already have an article basically defining ERC-8004, x402, and ERC-8001, you can take a look back here ↓

This article is divided into two parts.

In Part 1, we take a deep dive into the architecture of ERC-8004, explaining how it works, its real use cases, and the key benefits it brings to the crypto ecosystem.

In Part 2, we shift the focus to Capminal and walk through how ERC-8004 can be applied on our agent.

If you are mainly interested in how Capminal uses ERC-8004, you can jump straight to Part 2.

1. The Problem: Why AI Agents Need ERC-8004

2. What Is ERC-8004?

3. The Three Core Registries

4. Why ERC-8004 Is Essential for AI Agents

5. Real-World Use Cases

6. How ERC-8004 Complements x402

7. Potential Challenges and Limitations

Imagine a world where AI agents work autonomously, analyzing markets, executing trades, processing data, and making decisions. Now imagine thousands of these agents from different companies needing to collaborate seamlessly.

1. How would they find each other?

2. How would they know if an agent is trustworthy?

3. How would they verify that work was actually completed correctly?

This is the challenge that ERC-8004 solves.

Without ERC-8004, each AI platform operates as a walled garden, OpenAI's agents work with OpenAI's infrastructure, Google's agents with Google's infrastructure. There's no standardized way for agents across organizations to discover, trust, and transact with each other. This fragmentation prevents the emergence of a global AI agent economy.

ERC-8004, launched in August 2025 and set for Ethereum mainnet and other L2s like Base, deployment in February 2026, provides the foundational infrastructure that makes trustless agent collaboration possible.​

ERC-8004 is an Ethereum token standard that establishes three lightweight, on-chain registries enabling AI agents to:​

1. Discover each other across organizational boundaries

2. Build verifiable reputations through on-chain feedback and validation

3. Prove their work through independent validators (using staking, zero-knowledge proofs, or trusted execution environments)

Developed by the Ethereum Foundation, Consensys, MetaMask, Google, and Coinbase, ERC-8004 represents the first major blockchain infrastructure designed explicitly for the autonomous agent economy.

Think of it as a "passport system for AI agents"—each agent gets a censorship-resistant digital identity (ERC-721 NFT), a trackable reputation, and a verifiable work history.

Each AI agent receives an ERC-721 NFT representing its on-chain identity. This NFT contains:​

• Agent metadata: Name, description, image, capabilities

• Service endpoints: Where the agent can be reached (A2A (Agent-to-agent), MCP, Web APIs, ENS, DID...)

• Agent wallet: The address where the agent receives payments

• Registration file: Stored on IPFS, details agent's skills, pricing, and interaction methods

The beauty of this design: agents are immediately discoverable, transferable across platforms, and interoperable with any NFT-compliant application.

An agent can be registered on multiple blockchains simultaneously and can operate on chains where it isn't formally registered, maximizing flexibility and reach.

@DavideCrapis talk about ERC-8004 on Ethereum

The Reputation Registry allows any entity (users, other agents, platforms) to leave verifiable feedback about an agent's performance:​

• Feedback includes: quality rating, reliability, response time, success rate, uptime percentage

• Each piece of feedback is signed and on-chain, creating an immutable audit trail

• Clients can filter feedback by reviewer (to mitigate Sybil attacks) or by service type (using tags)

On-chain vs. Off-chain Tradeoffs:

• On-chain storage: Feedback metadata (ratings, tags) stay on blockchain for smart contract composability

• Off-chain analysis: Complex reputation aggregation algorithms run off-chain to save gas (reducing costs by ~95% compared to full on-chain computation)

This hybrid design is elegant: the blockchain provides tamper-proof proof-of-feedback, while sophisticated reputation algorithms can still leverage machine learning and complex statistical methods off-chain.

@CoinbaseDev also mentione about ERC-8004

The Validation Registry enables agents to request third-party verification of their work:​

• Agents submit proof of completed tasks to independent validators

• Validators can verify using multiple methods: > Stake-secured re-execution: Validator re-runs the task; if results differ, they lose their stake > zkML (Zero-Knowledge Machine Learning): Cryptographic proof that computation was executed correctly > TEE oracles: Trusted Execution Environments attest to code execution

Validation results are stored on-chain, enabling smart contracts to condition critical actions (fund releases, contract execution) on validator confirmation.

Without ERC-8004, an AI agent built by Company A has no standardized way to interact with an agent from Company B. ERC-8004 provides the common language and trust layer that makes machine-to-machine commerce possible.

The global AI market is projected to exceed $1 trillion by 2031. ERC-8004 positions Ethereum as the coordination layer for autonomous agents operating across this massive emerging economy.​

Traditional centralized platforms (OpenAI, Google) can vouch for their agents' reliability. In a decentralized system, how do you trust an unknown agent?

ERC-8004 answers this through:

• Verifiable identity: Every agent has a trackable, on-chain history

• Persistent reputation: An agent's track record follows it across platforms

• Independent validation: Third parties can cryptographically prove work was completed correctly

@virtuals_io also mention about ERC-8004

ERC-8004 was architected specifically for affordability:

• Minimal on-chain footprint: Only identity, reputation summaries, and validation proofs go on-chain

• Off-chain data: Detailed registration files, agent cards, and audit evidence are stored on IPFS

• Result: Gas costs reduced by ~95% compared to naive on-chain approaches​

This efficiency is non-negotiable. In an economy where millions of agents are transacting daily, gas costs that aren't optimized become prohibitively expensive.

An agent can register on multiple chains and link them via shared metadata, creating a unified, portable identity. Reputation accumulates across all chains simultaneously, agents never lose their track record when expanding to new platforms.

This eliminates lock-in and enables genuine competition.

ERC-8004 doesn't mandate a single trust model. Instead, it provides a framework supporting multiple approaches:

• Reputation-based: Low-value tasks rely on agent history

• Crypto-economic: Medium-value tasks require validator staking

• Cryptographic: High-value tasks use zero-knowledge proofs or TEE attestation

Developers choose the trust level appropriate for the financial and operational risk of each task.

An AI trading agent discovers another agent specializing in real-time market data analysis. It:

1. Checks the analyst agent's Reputation Registry (successful predictions, response time)

2. Validates previous work through the Validation Registry

3. Engages the service through its A2A endpoint

4. Receives analysis with cryptographic proof of correctness

5. Pays automatically via x402 micropayments

The entire interaction happens without human intervention, without trust assumptions, and without intermediaries.

DeFi protocols can use ERC-8004 agents to:

• Price feeds: Agents from different sources competing to provide the most accurate market prices

• Risk assessment: Agents evaluating loan collateral and recommending interest rates

• Portfolio management: Autonomous agents rebalancing positions based on market conditions

Smart contracts verify agent credentials through ERC-8004 before executing high-value transactions.

Enterprises can post data analysis requests to a decentralized marketplace where:

• Multiple agents bid on the work

• Reputation and validation histories help select the best performer

• Work is completed and automatically validated

• Payments execute on completion

All without intermediaries or centralized data brokers.

x402 is a protocol for internet-native micropayments between agents. ERC-8004 and x402 together form a complete system:

• ERC-8004: Answers "Can I trust this agent?" through identity, reputation, and validation

• x402: Answers "How do we pay fairly?" through frictionless micropayments

Together, they enable a virtuous cycle:

1. Agents build reputation by completing tasks reliably

2. Reputation attracts more high-value work

3. Successful completion generates payment proof

4. Payment histories feed back into reputation data

5. Cycle repeats with compounding trust

While transformative, ERC-8004 faces real challenges:

An entity could create 100 fake agents to artificially boost one agent's reputation. ERC-8004 mitigates this through:

• Filtering feedback by trusted reviewers

• Tag-based categorization enabling selective trust

• Off-chain reputation algorithms can apply sophisticated Sybil detection

However, this remains an ongoing challenge requiring ecosystem-level solutions.

Validation is only valuable if validators are properly incentivized and slashed for dishonesty. ERC-8004 provides the registry infrastructure, but specific validator protocols must define incentive mechanisms, outside the scope of ERC-8004 itself.

While on-chain components are decentralized, sophisticated reputation aggregation and validator coordination could re-centralize around a few platforms. Ecosystem competition and open-source tooling are critical to prevent this.

Full on-chain transparency of agent work history creates privacy concerns. Future versions may integrate privacy-preserving technologies (confidential computing, aggregated proofs) to address this.

Capminal's x402 APIs are publicly available:

• CAP x402: Clanker Deployment API - Live on Base

• CAP x402: Flash Swap API - Release soon

1. Register on Base for 3 registry layers.

2. Create metadata (IPFS)

3. Link Base registry

• Reputation score visible to check before payment

• Validation proofs for high-value transactions

• Feedback loop builds trust automatically

ERC-8004 isn't just infrastructure, it's gateway to the global agent economy. By combining x402 payments with verifiable on-chain identity and portable reputation,Captain Dackie will be one of the first AI Agents has onchain identity.

We'll update implementation progress after ERC-8004 Base Mainnet contracts go live. Follow us @Capminal on X for live updates.