Vibe coding is picking up momentum in WordPress agency circles, and Kinsta’s latest tutorial makes a strong case for why. The premise is straightforward: instead of writing code line by line, you work at the level of intent, defining architecture, logic, and business requirements, then delegating the actual code generation to an AI agent.
The practical payoff for agencies is meaningful. Automating repetitive tasks such as spinning up new client sites reduces time-to-market and keeps overhead from growing as the client roster expands. Onboarding a new project no longer has to mean proportionally more work for the development team.
What Vibe Coding Actually Requires
The article is careful to temper the enthusiasm. Vibe coding is not a shortcut that produces clean, maintainable code on autopilot. A solid grasp of programming fundamentals is still necessary. The shift is in how that knowledge is applied: toward guiding and verifying AI output rather than writing every statement by hand. Understanding how AI agents are architected, and how to prompt them reliably, becomes a core skill alongside traditional development knowledge.
The Workflow: Cloning a Blueprint Site
The tutorial centers on a real-world scenario: cloning a preconfigured WordPress blueprint site hosted on Kinsta using the Kinsta API and GitHub Copilot inside Visual Studio Code. The blueprint site used in the example included the Astra, GeneratePress, and Neve themes, along with Yoast SEO and Starter Templates.
To wire up the automation, you need three pieces of information from your Kinsta account:
- A Kinsta API key
- Your company ID (available in the MyKinsta dashboard URL)
- The environment ID of the site you want to clone (found under Sites, then the site’s Info section)
Why GitHub Copilot in VS Code
The author chose GitHub Copilot paired with VS Code for a few practical reasons. It is already embedded in the workflow of many agencies, so there is no learning curve around a new tool. Copilot has direct access to the project file system, letting it create and modify files based on prompts. It also supports a .github/copilot-instructions.md configuration file where you can define project-level rules the agent will follow consistently across the session.
Google AI Studio was used alongside VS Code as a prompt-building sandbox, useful for testing instructions in isolation before committing them to the actual project.
Copilot Modes Worth Knowing
The tutorial introduces Copilot’s three interaction modes. Plan mode generates a structured breakdown of tasks before any code is written, giving you a chance to review and adjust the roadmap. Once the plan looks right, execution is handed off to Agent mode. This separation keeps the agent focused on stated goals and makes it easier to track progress through a multi-step automation task.
For agencies evaluating workflow automation, this tutorial offers a grounded, tool-specific starting point rather than a theoretical overview, which makes it worth reading alongside any broader exploration of AI-assisted development.