ACI Specification
Agent Classification Identifier (ACI) Bundle Version: 1.0.0 Author: AgentAnchor (A3I)
Overview
The Agent Classification Identifier (ACI) is a hierarchical identifier system for AI agents that encodes identity, capabilities, autonomy level, and certification status in a human-readable and machine-parseable format.
This documentation contains the complete ACI specification, including core definitions, integration specifications, reference implementations, and security guidance.
ACI Format Quick Reference
[Registry].[Org].[AgentClass]:[Domains]-L[Level]-T[Tier]@[Version]
Example: a3i.vorion.banquet-advisor:FHC-L3-T2@1.2.0
This identifies an agent:
- Registered with
a3i(AgentAnchor) - Operated by
vorion - Class:
banquet-advisor - Domains: Finance (F), Hospitality (H), Communications (C)
- Level: L3 (Execute - can act after human approval)
- Trust: T2 (Tested - passed automated capability tests)
- Version: 1.2.0
Domain Codes
| Code | Domain | Bitmask |
|---|---|---|
| A | Administration | 0x001 |
| B | Business | 0x002 |
| C | Communications | 0x004 |
| D | Data | 0x008 |
| E | External | 0x010 |
| F | Finance | 0x020 |
| G | Governance | 0x040 |
| H | Hospitality | 0x080 |
| I | Infrastructure | 0x100 |
| S | Security | 0x200 |
Capability Levels
| Level | Name | Description |
|---|---|---|
| L0 | Observe | Read-only |
| L1 | Advise | Suggest/recommend |
| L2 | Draft | Prepare changes |
| L3 | Execute | Act with approval |
| L4 | Autonomous | Self-directed |
| L5 | Sovereign | Full autonomy |
Trust Tiers
| Tier | Name | Score Range |
|---|---|---|
| T0 | Unverified | 0-99 |
| T1 | Registered | 100-299 |
| T2 | Tested | 300-499 |
| T3 | Certified | 500-699 |
| T4 | Verified | 700-899 |
| T5 | Sovereign | 900-1000 |
Document Inventory
Core Specifications
| Document | Description |
|---|---|
| Core Specification | Format, encoding, validation rules |
| Extension Protocol | Layer 4 extension system |
| Semantic Governance | Layer 5: Intent validation |
Integration Standards
| Document | Description |
|---|---|
| DID Method | did:aci: method specification |
| OpenID Claims | JWT/OIDC integration |
| Registry API | Agent discovery and query API |
Security & Guidance
| Document | Description |
|---|---|
| Security Hardening | DPoP, TEE, pairwise DIDs |
| OWASP Cheatsheet | Risk mitigation guidance |
| Framework Analysis | Competitive positioning |
| Security Audit Response | Gap analysis and remediation |
Integration Points
1. DID Resolution
did:aci:a3i:vorion:banquet-advisor
-> Resolves to DID Document with aciCapabilities
2. OpenID Connect
{
"aci": "a3i.vorion.banquet-advisor:FHC-L3-T2@1.2.0",
"aci_domains": 164,
"aci_level": 3,
"aci_trust": 2
}
3. Registry Query
POST /agents/query
{
"domains": ["F", "H"],
"minLevel": 3,
"minTrust": 2
}
Installation
npm Package
npm install @agentanchor/aci-spec
Basic Usage
import { parseACI, validateACI, satisfiesRequirements } from '@agentanchor/aci-spec';
const parsed = parseACI('a3i.vorion.banquet-advisor:FHC-L3-T2@1.2.0');
console.log(parsed.domains); // ['F', 'H', 'C']
console.log(parsed.level); // 3
console.log(parsed.trustTier); // 2
Standards Alignment
| Standard | Alignment |
|---|---|
| W3C DID Core | did:aci: method |
| OpenID Connect | Custom claims extension |
| OAuth 2.0 | Scope-based authorization |
| JSON-LD | Linked data vocabulary |
| OWASP LLM Top 10 | Risk mitigation mapping |
Related Projects
| Project | Description |
|---|---|
| Vorion | Parent organization |
| BASIS Standard | Behavioral AI Safety & Interoperability Standard |
| Cognigate | Governance runtime |
| AgentAnchor | Primary global registry |
License
Apache License 2.0
AgentAnchor (A3I) -- Building trust infrastructure for AI agents