# How Kongctl Enables Deterministic, Declarative Platform Operations

Rick Spurgeon of Kong explains how Kongctl is designed to combine AI agents with deterministic platform operations by using declarative configuration, strong schemas, and auditable plans. His demo bootstraps a Kong AI gateway connected to OpenAI, generates a local Docker setup and certificates, applies changes through Kongctl, and routes a successful chat request through the gateway without hand-writing configuration. He argues that agent-driven workflows stay safer and more repeatable when the CLI exposes machine-friendly capabilities such as JSON-schema-based explain output, scaffolded YAML examples, built-in secret handling, and plan files that capture the exact diff to review, store, analyze, and apply. He then shows how an agent can create a GitHub-based CI/CD workflow that stages plans on pull requests, applies approved changes on merge, checks for drift, and updates the gateway by adding a second model. He also notes Kongctl supports federated configuration, broad but not complete Kong resource coverage, and coexists with Terraform and Kubernetes-based approaches.