Glossary

Governed Agentic Commerce Glossary

Extractable, citable definitions for the operating model behind governed commerce agents: GACO, PRISM, AVO, HarnexaHarness, AgentOps, CLEAR, risk classes, and the protocol stack.

A2A (Agent-to-Agent protocol)Protocol
An open protocol for secure communication and coordination between AI agents.
ACP (Agentic Commerce Protocol)Protocol
A commerce protocol for AI-assisted purchase flows where agents, people, and merchants complete orders together.
Agent Visibility Optimisation (AVO)AVO / GEO
Product-data readiness for AI buying agents that need to discover, compare, represent, and recommend products accurately.
AgentOpsOperations
The operating layer for deployed agents: identity, audit review, evaluation, cost, risk, and drift management.
Agentic commerceCommerce
Commerce workflows where AI agents discover, decide, propose, or transact within governed boundaries.
AP2 (Agent Payments Protocol)Protocol
A protocol direction for scoped, accountable payments initiated through agentic flows.
Append-only audit logGovernance
A database-backed event log where agent events are inserted but never updated or deleted.
Article 50 (EU AI Act)Regulation
EU AI Act transparency obligations for certain AI interactions and generated outputs.
Budget governanceGovernance
Per-agent limits on tool calls, tokens, time, or spend before a workflow can continue.
CLEAR frameworkEvaluation
HARNEXA's evaluation scorecard for Cost, Latency, Efficiency, Accuracy, and Reliability.
Commerce protocol stackProtocol
The emerging set of agent-readable commerce standards: MCP, WebMCP, A2A, UCP, ACP, AP2, and payment rails.
DESTRUCTIVE risk classRisk class
The risk class for irreversible tool actions such as deletion, credential revocation, or destructive system changes.
FINANCIAL risk classRisk class
The risk class for commercial, customer-impacting, or money-adjacent tool actions.
GACO FrameworkMethodology
HARNEXA's operating model for governed agentic commerce.
GEO (Generative Engine Optimisation)AVO / GEO
Content and structured-data work designed to be cited correctly by AI answer engines.
Governed agentGovernance
An AI agent constrained by identity, permissions, risk classes, approval gates, sensors, and audit logs.
HarnexaHarnessArchitecture
HARNEXA's governed execution loop for every agent tool call.
Human-in-the-loop (HITL)Governance
A workflow pattern where a human must approve before an agent crosses a consequential boundary.
LangGraphArchitecture
A graph-based framework for stateful, checkpointed agent workflows.
MCP (Model Context Protocol)Protocol
An open standard for connecting AI systems to tools and data sources.
Memory governanceGovernance
Rules for what agent memory can store, trust, retrieve, update, and forget.
ORACLEAgent
HARNEXA's specialist evaluation agent for CLEAR trajectory assessment.
PHANTOM VSAAgent
HARNEXA's flagship governed virtual sales assistant for CPG field visits.
PRISM readiness scoreMethodology
HARNEXA's 0-100 readiness score for governed agentic commerce workflows.
READ_ONLY risk classRisk class
The risk class for analysis or retrieval tool calls that do not change external state.
Trajectory evaluationEvaluation
Evaluation of the full agent run, including tool calls, risk classes, approval events, and outcomes.
UCP (Universal Commerce Protocol)Protocol
A commerce protocol direction for turning AI interactions into direct buying while merchants retain control.
WebMCPProtocol
A browser-context approach for exposing website actions as structured tools for AI agents.
x402Protocol
A payment protocol pattern that revisits HTTP 402-style payment flows for machine-readable transactions.