24.2 Using APEXlang with AI Coding Agents

Use APEXlang skills, SQLcl validation, and blueprints to collaborate with AI coding agents.

APEXlang is well-structured and easy to understand, so you can also work productively with AI coding agents like Claude, Codex, and others. With Oracle's APEXlang skills from GitHub, you can describe what you need and let an AI collaborator help review an application, suggest improvements, generate pages and components, make incremental changes, and find and fix problems using SQLcl's apex validate command.

AI agents can generate and update APEXlang to make any change represented in the declarative definition of an APEX application. With applications represented in a readable, editable format, you can describe intended changes, inspect the proposed source edits, validate them, and import the result back into APEX.

With AI agents generating and updating APEXlang, you can use natural language to modify APEX application artifacts including:

• Application attributes, settings, and metadata
• Pages, dialogs, regions, items, buttons, dynamic actions, and other UI components
• Supporting Objects, including data model definitions and installation scripts
• Shared Components, including themes, templates, lists, navigation, authorization schemes, and authentication schemes
• Business rules, computations, formulas, conditions, and validations
• Processes, workflows, automations, and other application logic
• Integrations, including REST Data Sources, OData Services, AI Providers, and Identity Providers
• Application, page, and component documentation, comments, and version information
• All other attributes, properties, and component settings exposed in App Builder

This workflow is most useful when you treat validation as part of the editing loop: change the APEXlang files, validate the application, review the results, and import only after the source validates cleanly.

Tip:

To install or update the APEXlang skills, use the skills sync command in SQLcl.

If you work with stakeholders to capture and agree on specifications before starting a project, APEX blueprints help teams refine requirements before costly surprises appear. Starting from a draft functional spec and data model description, an AI coding agent uses Oracle's detailed blueprint generation prompt to produce a structured Markdown plan for pages, navigation, reports, forms, charts, filters, actions, and initial behavior. App Builder imports this blueprint and deterministically generates a scaffolded app from tested APEXlang building blocks, letting stakeholders try real pages and give early feedback. You can revise the specification and regenerate until everyone agrees on the app’s shape. After sign-off, the final scaffold becomes the foundation you and your team evolve to add the details real systems require using a combination of App Builder and conversational development with AI coding agents using APEXlang skills.