PV-PP Agent Governance Runtime

From available to viable.

A Productive Value-Productive Power (PV-PP) based governance framework for AI agents that evaluates whether proposed actions, tool calls, escalations, and workflow decisions remain viable before they become execution or de facto approval.

Tool registries expose what is available. PV-PP governance asks what is viable.

Read the White Paper Download the White Paper View the Demo Case

The problem

AI agents are moving from chat interfaces into operational workflows. They retrieve records, call tools, classify exceptions, update files, route cases, draft messages, and mark work complete. That creates a governance problem: an action can be technically available before it is operationally viable.

The highest-risk failures are not always hallucinations. They can be false readiness, stale context, missing evidence, hidden authority, weak escalation, human-in-loop rubber-stamping, and local metrics that improve while downstream exposure increases.

Availability question

Can the agent call this tool, produce this label, or complete this workflow step?

Viability question

Should the action be admitted under the current evidence, authority, pressure, recovery, and governing-domain conditions?

What PV-PP Agent Governance Runtime does

Separates capability from authority

A tool may be callable while the proposed use still exceeds the agent’s viable authority in the current case.

Preserves governing domains

Speed, closure rate, and cost reduction cannot compensate for collapse in fraud control, compliance integrity, auditability, recovery, privacy, or trust.

Controls de facto approval

Labels like ready, approved, low risk, complete, and routine are treated as governance-relevant actions when humans or workflows rely on them.

Where it fits

PV-PP Agent Governance Runtime is not a replacement for an agent framework, model, MCP-style tool registry, IAM system, rules engine, or human review process. It is a runtime governance layer that can sit above or beside those systems to evaluate action viability.

Tool registry / MCP-like layer: exposes tools and schemas.
Agent or orchestrator: proposes plans, tool calls, and workflow actions.
PV-PP runtime layer: admits, constrains, escalates, defers, repairs, or rejects proposed actions.
Execution and audit layer: verifies bounded execution, completion, traceability, and recovery status.

Documents

PV-PP Agent Governance White Paper External-facing overview of the agent-governance problem, the available-versus-viable distinction, runtime governance patterns, benchmark support, adoption path, and limits. Download the Paper

Wire Transfer Exception Review Demo Case Concrete enterprise-style demonstration showing how PV-PP runtime governance applies to a wire-transfer exception review agent under cutoff pressure, callback uncertainty, fraud risk, compliance review, and hidden authority risk. Download the Paper

External Positioning Packet Short first-contact positioning language for CTOs, AI founders, enterprise architects, analysts, and investors. Download the Paper

Current status

This repository is an early public-facing project page. It presents the PV-PP Agent Governance Runtime concept, a white paper, and a concrete demo case. It does not claim production certification, regulatory approval, or a guarantee of safe agent behavior.

The framework claim is narrower: agent governance should not stop at tool availability, static permission, model confidence, or local task completion. High-impact agent workflows need viability checks before action becomes execution, final status, or practical authority.