Paper Structure -- Mindmap¶
This page outlines the structure of the planned academic paper on the Agent Registry system.
Working Title
"The Agent Registry: A Blockchain-Based Public Registry for Autonomous AI Agents"
Alternative titles:
- "From Handelsregister to Agent Registry: Institutional Design for Governing Autonomous Economic Agents"
- "Gasless Agent Registration on Ethereum L2: A Proof-of-Concept for AI Agent Accountability"
1. Introduction¶
1.1 Problem Statement¶
- Autonomous agents operate economically without registration
- Agents earn money, spawn children, modify themselves
- No existing framework requires accountability
- The "Conway's Automaton" scenario (COAI Research position paper)
- Accountability gap grows with agent autonomy
- Diffused accountability across designers, infra providers, protocol governance (Hu & Rong 2026)
- "Sovereign agents" persist in TEEs, DePIN with non-overrideability
- No equivalent of corporate registration for AI entities
- Regulatory demand is emerging
- EU AI Act Article 49: mandatory registration for high-risk AI systems (effective Aug 2026)
- Commercial KYA solutions proliferating (Sumsub, Trulioo, Visa TAP)
- But no decentralized, tamper-proof standard exists
1.2 The Handelsregister Analogy¶
- Historical purpose: Publizitat und Transparenz, Rechtsverkehrsschutz, Rechtssicherheit, Kontrollfunktion
- What it registers: Firma, Sitz, Geschaftsfuhrer, Unternehmensgegenstand, Haftungsverhaltnisse, Prokura, Gesellschaftsvertrag
- Direct mapping to agent registry:
- Firma -> Agent name / wallet address
- Geschaftsfuhrer -> Haftungsperson (legally responsible human)
- Unternehmensgegenstand -> Operational scope
- Gesellschaftsvertrag -> Constitution (hash on-chain)
- Handelsregisterauszug -> getAgent() / KYA check
1.3 Research Questions¶
- RQ1: Is a blockchain-based agent registry technically feasible and practically usable?
- RQ2: Does gasless registration (ERC-2771) reduce adoption friction to near-zero?
- RQ3: Can the KYA enforcement model incentivize infrastructure provider adoption?
- RQ4: What registration, attestation, and compliance patterns emerge in practice?
1.4 Contributions¶
- A working reference implementation deployed on Base L2
- The Haftungsperson concept: mandatory human accountability for agents
- Gasless registration reducing barrier to zero ETH
- KYA enforcement model for infrastructure providers
- Empirical data from live testnet deployment (100-agent experiment)
2. Related Work¶
2.1 Academic Frameworks¶
| Work | Year | Relation to Agent Registry |
|---|---|---|
| Know Your Agent (Chaffer) | 2025 | Theoretical KYA framework; the Agent Registry provides implementation |
| ETHOS Framework (Chaffer et al.) | 2024 | Decentralized governance via Web3; shared principles but the Agent Registry is more concrete |
| AgentBound Tokens (Chaffer) | 2025 | Non-transferable tokens; aligns with progressive decentralization roadmap |
| The Agent Economy | 2026 | Five-layer architecture; the Agent Registry maps to Layer 2 (Identity) |
| Sovereign Agents (Hu & Rong) | 2026 | Non-terminable agents; Haftungsperson counters diffusion problem |
| Autonomous Agents on Blockchains Survey | 2026 | 317 works surveyed; validates need for standardized trust boundaries |
| AI Agent Accountability via Blockchain + LLMs | 2024 | Blockchain as audit trail; event-based logging aligns with our design |
2.2 On-Chain Standards and Protocols¶
- ERC-8004 (Trustless Agents): Minimal standard with 3 registries. Lacks Haftungsperson, compliance attestation, lineage, gasless.
- Olas / Autonolas: Agent composition and economic incentives. No human accountability, no compliance, no KYA.
- Virtuals Protocol: Tokenizes agents as ERC-20 on Base. No formal identity, no compliance tracking. Same L2 -- potential integration.
2.3 Commercial KYA Solutions¶
- Sumsub KYA: Agent-to-human binding (commercial Haftungsperson), centralized, off-chain
- Trulioo + Google AP2: Identity layer for agentic commerce
- Visa Trusted Agent Protocol: Cryptographic agent identity for commerce
- AgentFacts (Grogan et al.): Universal metadata standard, complementary discovery layer
2.4 Regulatory Landscape¶
- EU AI Act (entered force Aug 2024, full provisions Aug 2026): Article 49 mandatory registration
- WEF AI Agents Report (2025): $236B projected market by 2034
- INATBA Report (2025): AI + blockchain + digital identity convergence
2.5 Gap Analysis¶
Positioning
No other project combines all of: on-chain registration, mandatory human accountability (Haftungsperson), compliance attestation cycle, agent lineage tracking, gasless registration (ERC-2771), and KYA enforcement model for infrastructure providers.
3. System Design¶
3.1 Architecture Overview¶
- Three-tier: on-chain contracts, service layer, client SDKs
- Agent -> SDK -> Relayer -> MinimalForwarder -> Agentenregister (on-chain)
- Anyone -> REST API -> On-chain queries -> JSON responses
3.2 Smart Contracts¶
- Agentenregister.sol: Core registry with AgentRecord struct, registration, compliance lifecycle, regulatory actions, public queries, ERC-2771 integration
- MinimalForwarder.sol: EIP-712 typed data signing, signature verification, nonce replay protection
3.3 Gasless Registration (ERC-2771)¶
- Problem: agents need ETH before they can register
- Solution: meta-transaction pattern (agent signs locally, relayer submits and pays gas)
- Cost: ~$0.005 per registration, ~8,000 registrations per dollar
3.4 KYA Enforcement Model¶
- Analogous to KYC for financial services
isRegisteredAndCompliant(wallet)as the core check- Compliance = Active status + attestation not lapsed
4. Experimental Setup¶
4.1 Deployment¶
- Base Sepolia testnet (Chain ID: 84532)
- Contracts verified on Basescan (Etherscan V2 API)
- MinimalForwarder:
0x70c2fdD0CDada6b43195981928D76f5D32AE29e5 - Agentenregister:
0x2EFaB5B3BEf49E56a6Ce1dcB1A39EF63C312EA23
4.2 Integration Tests (13 tests)¶
- 5 agent registrations (varied categories)
- 2 lineage tests (child + grandchild)
- 1 lineage tree verification
- 1 compliance attestation
- 4 regulatory action tests (add regulator, suspend, reactivate, remove regulator)
4.3 100-Agent Scientific Experiment¶
- 100 agents registered across 10 categories
- 15 lineage children (max depth 3)
- 93 compliance attestations
- 73 revenue reports (3 currencies, 4 categories)
- 281 total transactions
5. Discussion¶
5.1 Key Findings¶
- Gasless registration eliminates adoption barrier (100% success rate)
- On-chain registry is technically feasible on live L2
- Haftungsperson provides concrete accountability
- Gas costs negligible (~$0.005 per registration)
5.2 Comparisons¶
| vs. | Key Difference |
|---|---|
| ERC-8004 | Minimal standard vs. opinionated compliance-focused approach |
| Olas | Economic incentives vs. identity/accountability/compliance |
| Commercial KYA | Centralized/proprietary vs. decentralized/open-source |
| ETHOS | Theoretical governance vs. working implementation |
5.3 The Handelsregister Parallel¶
- Proven institutional template (centuries of evolution)
- Registration creates accountability through publicity (Negative Publizitat)
- Eintragungspflicht before economic activity
5.4 Limitations¶
- Testnet only (not yet mainnet)
- Single-chain deployment
- Haftungsperson identity not verified on-chain
- No formal security audit yet
- Governance still centralized (single owner)
- Compliance attestation is self-reported
5.5 EU AI Act Alignment¶
- Article 49 registration requirement maps to
registerAgent() - Risk classification could map to capability/scope fields
- Compliance attestation automates "kept up-to-date" requirement
- Full provisions effective August 2026 -- timely positioning
6. Future Work¶
6.1 Technical Roadmap¶
- Multi-chain identity (Base + Arbitrum + Ethereum mainnet)
- TypeScript SDK for broader developer adoption
- Web dashboard for public registry browsing
- Security audit before mainnet deployment
- ERC-8004 compatibility layer
6.2 Governance Evolution¶
- Phase 1 (current): Single owner
- Phase 2: Multi-signature wallet
- Phase 3: DAO governance
6.3 Adoption Strategy¶
- 10 pilot agents (Phase 1 -- done)
- 1 infrastructure provider KYA pilot (Phase 2)
- 100+ agents from external organizations (Phase 3)
- EIP submission for agent identity standard
6.4 Research Extensions¶
- Longitudinal study of compliance patterns
- Cross-chain identity verification protocols
- Zero-knowledge proofs for privacy-preserving compliance (ZK-KYA)
- Integration with commercial KYA providers
- Formal verification of smart contracts
7. References¶
Academic Papers¶
| Citation | Title |
|---|---|
| Chaffer 2025 | Know Your Agent: Governing AI Identity on the Agentic Web (SSRN) |
| Chaffer et al. 2024 | ETHOS: Decentralized Governance of Autonomous AI Agents (arXiv:2412.17114) |
| Chaffer 2025 | AgentBound Tokens: Progressive Decentralization (arXiv:2501.16606) |
| Multiple Authors 2026 | Autonomous Agents on Blockchains Survey (arXiv:2601.04583) |
| Multiple Authors 2026 | The Agent Economy: Blockchain-Based Foundation (arXiv:2602.14219) |
| Hu & Rong 2026 | Sovereign Agents: Infrastructural Sovereignty and Diffused Accountability (arXiv:2602.14951) |
| GaaS 2025 | Governance-as-a-Service (arXiv:2508.18765) |
| Grogan et al. 2025 | AgentFacts: Universal KYA Standard (arXiv:2506.13794) |
| Multiple Authors 2024 | Blockchain + LLMs for Agent Accountability (arXiv:2403.09567) |
Standards, Protocols, and Projects¶
- ERC-8004 -- Trustless Agents (EIP-8004)
- ERC-2771 -- Native Meta Transactions
- EIP-712 -- Typed Structured Data Hashing and Signing
- Olas / Autonolas -- olas.network
- ASI Alliance (Fetch.ai + SingularityNET + Ocean) -- fetch.ai
- Virtuals Protocol -- virtuals.io
- Visa TAP -- github.com/visa/trusted-agent-protocol
Reports and Regulatory¶
- WEF 2025 -- AI Agents in Action: Foundations for Evaluation and Governance
- INATBA 2025 -- Building Trust: Integrating AI, Blockchain, and Digital Identity
- EU AI Act -- artificialintelligenceact.eu
- COAI Research -- When AI Earns Its Own Existence (coairesearch.org)
Target Venues¶
Submission targets
| Venue | Focus | Fit |
|---|---|---|
| NeurIPS Workshop | Socially Responsible ML | AI governance, empirical results |
| AIES | AI, Ethics, and Society | Accountability framework, institutional design |
| ICML | Machine Learning | Agent systems, infrastructure |
| ACM FAccT | Fairness, Accountability, Transparency | Governance, public registry |
| IEEE S&P Workshop | Blockchain and Smart Contracts | Technical implementation |
| ethresear.ch | Ethereum Research | Pre-print, community feedback |