Building the independent evidence layer for the agentic economy.

AI systems are moving from generating answers to taking actions.

Agents can now initiate payments, call APIs, modify systems, coordinate with other agents and make decisions across infrastructure owned by different organisations. But as autonomy increases, the ability to independently verify what happened, why it happened and whether an action should be trusted has not kept pace.

INEVRI is developing APP, the Agentic Provenance Protocol: an open, vendor-neutral standard for independently observing and verifying consequential autonomous activity.

APP creates an evidence-backed view of an action as it moves across agents, models, tools, APIs, services and execution environments. When an autonomous workflow crosses multiple independently controlled systems, APP correlates the available evidence into a defensible provenance chain without requiring any single vendor, model or platform to be the source of truth.

This provides organisations with evidence to answer fundamental questions:

• Who or what initiated the action?
• What authority did it have?
• What systems and information influenced it?
• What changed as the action crossed organisational boundaries?
• Was the resulting action consistent with policy and expected behaviour?
• And when the evidence is incomplete or contradictory, can the system recognise that uncertainty rather than manufacture confidence?

APP is designed for environments where autonomous activity can create material financial, operational, scientific or physical consequences, including financial infrastructure, enterprise systems, critical and sovereign infrastructure, scientific computing and physical autonomous systems.

The protocol follows a simple principle:

Protect first. Identify second. Trust progressively.

APP does not assume that an agent, platform or upstream system is trustworthy. It observes at system boundaries, preserves evidence, distinguishes what is verified from what is asserted or inferred, and enables enforcement at points of consequence.

INEVRI is developing APP as an open standard rather than another proprietary agent platform. It is designed to work alongside existing identity, security, observability and execution infrastructure while providing an independent evidence layer across them.

As autonomous systems become increasingly interconnected, trust cannot depend solely on the systems performing the action.

It needs independent evidence.