PROPOSED BY ANTHONY JAMES PEACOCK

AJP AI Agent Schema Standard

v0.1.0 — Draft

A formal vocabulary for describing AI agents in the knowledge graph. Defines 20 schema types covering agent identity, capability declaration, cryptographic verification, and deployment accountability.

AUTHOR
Anthony James Peacock
STEWARD
LinkDaddy LLC
FIRST IMPLEMENTATION
aiverified.io
LICENSE
CC BY 4.0
NAMESPACE
https://schema.anthonyjamespeacock.com/
PREFIX
ajp:
QUICK START
DOWNLOAD VOCAB
{
  "@context": {
    "@vocab": "https://schema.org/",
    "ajp": "https://schema.anthonyjamespeacock.com/"
  },
  "@type": ["SoftwareApplication", "ajp:AIAgent"],
  "ajp:agentPurpose": "...",
  "ajp:systemPromptHash": "sha256:...",
  "ajp:capabilityDeclaration": {
    "@type": "ajp:CapabilityDeclaration",
    "ajp:authorisedActions": [...],
    "ajp:prohibitedActions": [...]
  }
}
TYPES (20)
ajp:AIAgentextends schema:SoftwareApplication

An artificial intelligence agent — a software system designed to autonomously or semi-autonomously take actions, answer …

16 props →
ajp:AgentPassportextends schema:EducationalOccupationalCredential

A cryptographically sealed identity credential for an AI agent. The Agent Passport enables Know Your Agent (KYA) verific…

11 props →
ajp:CapabilityDeclarationextends schema:DigitalDocument

A structured document declaring the explicit capabilities and prohibitions of an AI agent. Cryptographically sealed so a…

7 props →
ajp:AuthorisedActionextends schema:Action

A single, explicitly permitted action that an AI agent may perform, with defined scope, constraints, and human oversight…

12 props →
ajp:ProhibitedActionextends schema:Thing

A single, explicitly forbidden action that an AI agent must never perform, with a stated reason and consequence of viola…

5 props →
ajp:TrainingDataSourceextends schema:CreativeWork

A source of knowledge used to train or configure an AI agent's responses. Provides provenance information so users can e…

7 props →
ajp:ModelConfigurationextends schema:Thing

The technical configuration of the AI model(s) powering an agent, enabling transparency about the underlying technology …

10 props →
ajp:AgentDeploymentextends schema:Thing

A record of a specific AI agent being deployed on a specific domain by a specific organisation. Creates the accountabili…

9 props →
ajp:HumanApprovalGateextends schema:Thing

A defined condition under which an AI agent must pause execution and obtain explicit human approval before proceeding. I…

8 props →
ajp:AgentPersonaextends schema:Thing

The presented identity of an AI agent as experienced by users. Separate from the underlying AIAgent to allow branded dep…

6 props →
ajp:KnowledgeBoundaryextends schema:Thing

A declaration of the scope and limits of an AI agent's knowledge — what domains it covers, what it explicitly excludes, …

6 props →
ajp:AgentAuditRecordextends schema:DigitalDocument

A record or pointer to the audit log of an AI agent's actions. Provides accountability evidence that the agent has opera…

7 props →
ajp:DeploymentConstraintsextends schema:Thing

A set of constraints defining the acceptable operational parameters for deploying an AI agent, including geographic, sec…

8 props →
ajp:AgentIncidentextends schema:Event

A formal record of an incident where an AI agent behaved outside its declared capability boundaries or caused harm. The …

16 props →
ajp:AgentVersionextends schema:CreativeWork

A specific version of an AI agent within a version chain. Links previous and successor versions by hash, enabling tamper…

12 props →
ajp:AgentLicenceextends schema:DigitalDocument

Commercial and usage licensing terms for an AI agent. Machine-readable for procurement systems, comparison engines, and …

13 props →
ajp:AgentCollaborationextends schema:Thing

A structured description of how two or more AI agents work together within a multi-agent system. Defines the orchestrati…

13 props →
ajp:AgentEndorsementextends schema:Review

A formal attestation by a third party that an AI agent meets specified standards, has passed testing, or is approved for…

16 props →
ajp:AgentFeeextends schema:PriceSpecification

The pricing model declaration for an AI agent. Machine-readable for procurement systems, comparison engines, and AI assi…

17 props →
ajp:ConsentDeclarationextends schema:DigitalDocument

A machine-readable declaration of the data collected by an AI agent, the legal basis, retention period, and user rights.…

22 props →
MACHINE-READABLE ENDPOINTS

All endpoints support content negotiation. Add Accept: application/ld+json to get JSON-LD.

# Full vocabulary
GET https://schema.anthonyjamespeacock.com/
    Accept: application/ld+json

# Individual type
GET https://schema.anthonyjamespeacock.com/AIAgent
    Accept: application/ld+json

# Human-readable (default)
GET https://schema.anthonyjamespeacock.com/AIAgent
WHY THIS STANDARD EXISTS

Schema.org's SoftwareApplication type is inadequate for AI agents. It has no concept of capability boundaries, training provenance, cryptographic integrity, human approval gates, or deployment accountability chains.

As AI agents begin handling financial transactions, customer relationships, and regulated activities on behalf of businesses, the absence of standardised machine-readable identity creates a critical trust gap.

This vocabulary closes that gap. It is designed to be immediately usable, submitted to Schema.org as a formal community proposal, and adopted by any platform that operates AI agents.