`kongctl` was built with the modern development stack in mind. Developers working in a terminal want the ability to query systems quickly to verify state, behaviors, and capabilities. Coding harnesses need well defined tools and accurate schemas for efficient model generation of code and configuration. More than ever, software delivery pipelines require deterministic and auditable automations when changing the state of critical path systems, like API platforms.

`kongctl` brings these together in one tool — a fast, natively compiled CLI with a readable command grammar and a declarative engine underneath.

`kongctl` is not just for people on a keyboard. It's designed to be driven by AI coding agents natively.

That starts with bundled *agent skills* you can pull into a project in one command (`kongctl install skills`). These skills give coding harnesses ready-made instructions for using kongctl across a variety of tasks. We will continue to curate and deliver new skills while improving them over time with agent-led evaluations.

Beyond skills, `kongctl` has self-describing capabilities. The `kongctl explain <resource-type>` command generates the latest JSON-Schema resource shape for a given type. This allows a coding model to reliably generate conforming configuration without reaching out to online documentation or other help text, saving context and time. `kongctl scaffold <resource-type>` is similar but produces a complete self-documented YAML configuration starting point — useful for agents and humans alike.

`kongctl` outputs are just as flexible: commands can emit various formats depending on the need. JSON, YAML, text, and in some cases Helm and other formats are supported. Built-in `jq` filtering (no external tool required) gives your coding harness the ability to query values without deferring to another tool or passing the coding model unnecessary data. For example `kongctl get api my-api --jq '.description'` only retrieves the given API description field without any additional steps or tools increasing coding agent efficiency.

The everyday ergonomics are built for human users, too. Authorization is handled through a web-based authorization flow or token-based configuration. A single `kongctl login` gets you connected without dealing with tokens by hand.

From there, deterministic `kongctl <verb> <noun> <options> commands` start to feel natural — `kongctl get apis` reads like what it does. Named configuration profiles make switching between Konnect organizations or configurations easy. Readable and configurable terminal output, clear error messages, and an interactive TUI based resource viewer (`kongctl view`) helps you view the state of your API Platform without leaving your development environment.

When the built-in features aren't enough, `kongctl` supports locally installed extensions. Developers can build their own commands against the `kongctl` SDK and install them so they run with the same native look and feel as the rest of the CLI.

The headline capability in `kongctl` is a state-free declarative engine.

You define your API platform infrastructure as plain YAML, and let `kongctl` reconcile Konnect to match. There is no state file to store, lock, or drift — `kongctl` computes differences by querying the live state within the target organization.

The declarative engine employs a *plan-based* solution. This means `kongctl` calculates the difference between the input declarative configuration and the live resources. This difference is stored in a *plan* document. The plan is then easily reviewed, audited, transported, and optionally later used as the basis for applied changes to live resources.

This plan-based approach serves two audiences at once. It makes GitOps practical — generate a plan in a pull request, review it, and apply it on merge — while giving humans a real audit point before the changes land.

`kongctl` covers the core of Konnect: APIs, dev portals, application auth strategies, control planes, event gateways, and analytics dashboards. It also manages organization structure — teams, roles, system account and user assignments, and portal team group mappings for IdP integration.

The [_declarative configuration with kongctl guide_](https://developer.konghq.com/kongctl/declarative/)_declarative configuration with kongctl guide_ provides the full details for using this advanced capability.

Users familiar with Kong API Gateway declarative management tool [_decK_](https://developer.konghq.com/deck/)_decK_ may wonder how these tools are related.

decK remains Kong's tool for configuring Kong Gateway. kongctl operates one level up, at the Konnect API platform — APIs, portals, control planes, teams, and more. It's an expansion of the toolchain, not a replacement, and a Konnect-native alternative to managing those resources through Terraform.

But `kongctl` and decK have an [_integration point_](https://developer.konghq.com/kongctl/declarative/#resources-managed-by-deck)_integration point_. By declaring a `_deck` entry into a control plane's configuration, `kongctl` can invoke decK inline giving users a single command feel for API Gateway configuration.

There's a second AI story here: `kongctl` isn't just built *for* agents, it's built *with* them.

Alongside the usual coding-agent assistance, we run a fleet of agentic workflows against [_the open source kongctl repo_](https://github.com/Kong/kongctl)_the open source kongctl repo_ that provide ongoing services to the codebase, helping our team move faster and ship a higher quality tool. 

• - **A feature-user agent.** We deploy an agent as a `kongctl` *user* — not a developer — give it minimal guidance, and let it invent its own use cases. When it hits friction, it files an issue with the prompts and details, and that feedback flows back into the CLI, skills, and docs.
• - **End-to-end test-gap scanning.** `kongctl` runs 125+ e2e test scenarios and 2,000+ validations running against production Konnect organizations. A nightly agent scans that suite for coverage gaps and files issues with concrete suggestions — which a coding agent can then pick up with all the context it needs.
• - **Agentic issue triage.** New issues are evaluated automatically — labeled, summarized, and annotated with debugging strategies and links to relevant code and resources.
• - **Performance benchmarks and regression detection.** A nightly job runs a benchmark suite across configuration sizes and scans for regressions in elapsed time *and* API request counts. Watching request counts catches unintended behavior changes for the same inputs — and has helped surface real Konnect networking issues for engineering to resolve.

Ready to get started? Visit the Kong Developer site for the [_complete getting started guide_](https://developer.konghq.com/kongctl/get-started/)_complete getting started guide_. 

Three areas where you'll see `kongctl` grow quickly:

• - **Broader resource coverage.** Konnect is growing further into Identity, new runtimes, and AI use cases. `kongctl` will strive to maintain parity with the platform as a whole.
• - **Deeper docs integration.** Expect `kongctl` to show up across the [_Kong Developer_](https://developer.konghq.com/)_Kong Developer_ site — with getting-started guides, recipes, and workflow walkthroughs built around it.
• - **More agent capabilities.** New bundled skills, expanded `explain/scaffold `coverage, and additional CI/CD integrations are on the way.

Try the tool out and let us know what you think — [_GitHub_](https://github.com/Kong/kongctl)_GitHub_, [_Kong Nation_](https://discuss.konghq.com/)_Kong Nation_, or reach out to your Kong support resource.