Documentation

Getting Started

Understand ValidKernel architecture and evaluate integration with your environment.

Access Model

ValidKernel is currently available through controlled partner access. This page explains the architecture and evaluation paths available to you.

What is ValidKernel?

ValidKernel is deterministic trust infrastructure for enterprise AI systems. It operates as a stateless boundary validation layer that enforces governance policies with 0.01ms latency and 100% deterministic output.

The kernel does not negotiate. It does not adapt. It enforces. Same input, same output, every time.

Architecture Overview

L0

Human Governance

Ultimate authority. Sets policy and defines invariants.

L1

Kernel

Enforcement layer. Validates, enforces, and audits.

L2

Proposer

Request layer. Requests authorization within constraints.

Evaluation Paths

1. Try the Mirror (No Account Required)

The ValidKernel Mirror is a public query gateway that demonstrates governed interaction. Submit up to 25 queries per session with no sign-up required.

Try Mirror

2. Review the API Contract

The API Reference documents the endpoint contract surface. Review the validation, mirror-query, and audit endpoints to understand the integration model.

API Reference

3. Request Partner Access

For API credentials, SDK access, and direct engineering support, request partner access through our contact form. We will evaluate your use case and provision appropriate access.

Request Access

Core Capabilities

CapabilitySpecification
Validation Latency0.01ms
Determinism100%
Compression Ratio5,600:1
Benchmark Tests52/52 passed
DeploymentOn-premise or cloud
ComplianceISO 42001, EU AI Act, SOC 2

Next Steps