VEX
Governed Execution
agent governanceexecution controlaudit-ready authorization
Governed execution is the principle that no autonomous AI agent action should reach a production system without explicit policy evaluation, authorization context, and reviewable evidence.
WHY IT EXISTS
Autonomous agents can turn model output into real consequences: files changed, APIs called, tickets closed, purchases made, or infrastructure modified. Traditional prompt safety controls operate before the model speaks; governed execution controls the point where a proposed action becomes a real action.
PROTOCOL CONTEXT (VEX + MCP)
In ProvnAI's architecture, McpVanguard enforces real-time security policy at the MCP tool-call boundary, while VEX Protocol preserves authorization context, outcome, and witness evidence for actions that need governance and later review.
Proposal
The agent describes the action it intends to take.
Authorization
Policy, identity, scope, and intent are evaluated before execution.
Execution
The action proceeds only if the boundary permits it.
Evidence
The decision and outcome are preserved for audit, review, and incident response.
ProvnAI Mitigation
Governed execution reduces implicit trust in model reasoning. It makes policy enforcement external to the model and preserves evidence about what crossed the execution boundary, why it was allowed or denied, and what happened next.