Docs

OpenKedge documentation

The docs cover the control model, approval flow, and concrete examples that show how OpenKedge keeps important AI actions governed in production systems.

Concepts

Intent

Why AI actions begin as stated requests instead of direct execution.

Concepts

Contracts

How OpenKedge narrows approved actions into bounded execution contracts.

Concepts

IEEC

The evidence model that makes every important action explainable and auditable.

Architecture

Pipeline

The seven-stage architecture from intent proposal to evidence chain.

Examples

DevOps Guardrail

A concrete EC2 termination scenario showing policy rejection before execution.

Implementation framework

OpenKedge makes PDDS practical.

The framework translates the paradigm into protocol mechanics, reference implementations, evidence records, and governed runtime boundaries.

Intent

Context

Policy

Contract

Identity

Execution

Evidence

Replay

Governed action pipeline

Decision stages

Intent Proposer

The AI system states what it wants to do (its proposed intent) before it receives any operational authority.

Context Check

The system reviews situational factors, dependencies, risk profile, timing, and current operating conditions.

Policy Review

The proposed intent is evaluated against governing rules, boundaries, and organizational policies.

Execution Contract

Any permitted action is bound to a specific execution contract defining its scope, limits, and expiration.

Proof-Derived Execution Identity

The system issues a cryptographic, time-bounded identity valid only for the duration of the approved contract.

Controlled Action

Only the approved state transition is permitted to execute, with no expansion or hidden branching.

Evidence Chain / IEEC

Every step—from proposed intent to final outcome—is recorded as a cryptographically linked evidence chain.

Replay & Policy Refinement

The recorded evidence feeds back into replay, simulation, policy adjustment, and continuous protocol refinement.

Open source implementations

Reference code for governed autonomous infrastructure.

OpenKedge provides implementation patterns that teams can inspect, adapt, and use as a foundation for PDDS-aligned pilots.

1. Pick a high-impact scenario

Choose an AI action with real operational risk: cloud change, data workflow, service action, or approval process.

2. Map the authority boundary

Define who owns the rule, which approvals matter, what scope is allowed, and what evidence leadership needs.

3. Run the governed pilot

Test the action, watch policy respond, and review the evidence trail with executive, compliance, and technical teams.

Example

Clone GitHub

import { createProposal, evaluateProposal } from "@openkedge/sdk-js";

const proposal = createProposal({
  actor: "agent.ops.autoscaler",
  target: "aws:ec2:i-0ab1cdef23456789",
  intent: "terminate_instance",
  desiredOutcome: "remove_unused_capacity",
});

const decision = await evaluateProposal(proposal);

if (decision.status === "approved") {
  await decision.execute();
}