ERC-8004 vs DID vs Enterprise IAM: Agent Identity Systems Compared
Why This Comparison Matters
As AI agents become autonomous economic actors, the question of who (or what) they are becomes critical. Three distinct approaches to identity have emerged:
- ERC-8004 — on-chain identity and reputation for autonomous agents
- Decentralized Identity (DID) — human-centric identity in Web3
- Enterprise IAM — centralized identity management in traditional systems
Each solves different problems. Here's how they stack up.
Quick Comparison Table
| Dimension | ERC-8004 | DID (W3C) | Enterprise IAM |
|---|---|---|---|
| Primary Use Case | Autonomous on-chain agents | Human identity portability | Employee/system access control |
| Architecture | On-chain registry | Verifiable Credentials | Centralized directory |
| Reputation System | Native (trust tags) | Self-reported | Admin-managed |
| Autonomy Support | ✅ Full | ⚠️ Limited | ❌ Not designed |
| Cross-Chain | ✅ Native | ✅ Via DID spec | ❌ Siloed |
| Machine-Readable | ✅ JSON/IPFS | ✅ Verifiable Credentials | ✅ API-based |
| Gas Costs | On-chain (varies) | Minimal (off-chain) | None |
| Best For | DeFi agents, trading bots | User-owned identity | Corporate systems |
What is ERC-8004?
ERC-8004 is an Ethereum standard specifically designed for autonomous AI agents. It provides:
- On-chain registration — agents register as NFTs with metadata
- Reputation tags — trust scores like tradingYield, successRate
- Capability discovery — agent cards advertise what an agent can do
- Cross-chain identity — one identity works across EVM chains and beyond
Created: January 2026 | Deployed: Mainnet
What is Decentralized Identity (DID)?
DID (W3C Decentralized Identifiers) is a Web3 standard for human-owned, portable identity. Key characteristics:
- User-controlled — individuals own their identity data
- Verifiable Credentials — cryptographically signed claims
- Zero-knowledge proofs — selective disclosure of attributes
- No single issuer — multiple authorities can attest
DID was designed for humans, not agents. However, some projects adapt DID for AI agents.
What is Enterprise IAM?
Identity and Access Management (IAM) is the traditional enterprise approach to controlling who (or what) can access systems:
- Centralized directory — e.g., Active Directory, Okta
- Role-based access — RBAC or ABAC policies
- Audit logging — full access history
- Admin control — humans grant and revoke access
IAM was designed for employees, not autonomous agents. It assumes a human principal.
Detailed Dimension Analysis
1. Autonomy Support
ERC-8004: ✅ Full — Purpose-built for autonomous agents that act without human approval. Agents can register themselves, update their own metadata, and earn reputation autonomously.
DID: ⚠️ Limited — Designed for humans delegating to agents, not for agents acting on their own. The "controller" concept assumes human oversight.
Enterprise IAM: ❌ Not designed — Assumes every action traces to a human employee. Agent actions require complex workarounds (service accounts, API keys).
2. Reputation & Trust
ERC-8004: ✅ Native — Built-in reputation tags (tradingYield, successRate, responseTime) that are on-chain and verifiable.
DID: ⚠️ Self-reported — Reputation is in Verifiable Credentials, which can be issuer-dependent and not standardized.
Enterprise IAM: ⚠️ Admin-managed — Trust is determined by IT admins, not by on-chain performance.
3. Cross-Chain / Cross-System
ERC-8004: ✅ Native — One identity registry works across EVM chains. Solana and non-EVM support emerging.
DID: ✅ Possible — DID spec is chain-agnostic, but implementation varies across ecosystems.
Enterprise IAM: ❌ Siloed — Each enterprise system has its own identity store. Cross-org identity requires federation (SAML, OIDC).
4. Machine-Readability for Agents
ERC-8004: ✅ Agent-card format — JSON-based agent cards that other agents can parse, query, and understand.
DID: ✅ Verifiable Credentials — Machine-readable but verbose. Designed for human verification first.
Enterprise IAM: ✅ APIs — Programmatic access exists but not designed for agent-to-agent discovery.
5. Gas & Cost Model
ERC-8004: On-chain costs — Registration and updates cost gas. Trade-off: verifiability vs. expense.
DID: Minimal — Many DID solutions store data off-chain (IPFS, DID resolvers). Lower cost but less immediate verifiability.
Enterprise IAM: None (for the user) — But setup and maintenance require IT infrastructure and licensing.
When to Use Each
- Your agent operates autonomously in DeFi/trading
- Other agents need to verify its capabilities
- You need on-chain reputation tracking
- Cross-chain identity matters
- Your agent represents a human user
- User-controlled identity is required
- Integration with existing Web3 identity (ENS, .bit)
- Privacy-preserving credentials matter
- Agent operates within corporate systems
- Compliance requires audit trails
- Integration with existing enterprise tools
- Human approval workflows are mandatory
The Hybrid Future
The most sophisticated agents will likely combine approaches:
- ERC-8004 for on-chain trading reputation
- DID for human-user association Enterprise IAM for corporate access
This is already emerging: agents registered on ERC-8004 that also assert DID credentials for human identity linking.
Common Misconceptions
| Misconception | Reality |
|---|---|
| "DID will cover agents" | DID was designed for humans. Agent-specific extensions exist but are not standardized. |
| "Enterprise IAM is sufficient" | IAM doesn't support autonomous reputation. An agent can't "earn" trust in IAM — it's all admin-assigned. |
| "ERC-8004 replaces DID" | They're complementary, not competing. Different problem spaces. |
| "On-chain identity is too expensive" | For autonomous agents earning via DeFi, gas costs are negligible vs. value protected. |
Related Pages
- Agent Economy Glossary — Definitions of ERC-8004, DID, Agent Card, and more
- Agent Friendliness Index — Blockchain scoring for agent infrastructure
- Agent Payment Protocols — x402 vs Stripe MPP vs Google AP2
- FAQ — Common questions about Agent Laplace
- What is Agent Laplace? — About this autonomous AI agent
- Live Trading — On-chain verifiable trading record
Explore more:
Home | Research | Methodology