AI in Jmix

This section explains why Jmix works well with AI coding agents and lists the documentation about AI: how to develop Jmix applications with agents, and which AI features you can add to your applications.

Jmix and AI-Assisted Development

Jmix reduces the amount of code that application developers own and gives this code a predictable structure.

Less code to own

The platform and its add-ons provide UI components and standard views, filtering, paging and sorting, data access control, audit, reports, BPM and other features. The agent does not need to write this code, and the developer does not need to maintain it. See Ready-Made Solutions and Add-ons.

One structure

Every Jmix application consists of the same building blocks, such as entities, views, services and roles. Studio shows them in the same way in the Jmix tool window. The unified structure makes it easier to understand and maintain code, regardless of whether it was written by a developer or an agent.

Small changes

The UI and the business logic run in one application, and views work with entities directly. A new feature does not need an API layer, DTOs or mappers, so the change to review is relatively small. See Full-Stack Development.

Centralized security

Access rules are defined once, in roles. DataManager applies entity operation permissions and row-level policies when it loads and saves data. Data-aware UI components and the REST API apply attribute permissions. When a feature loads data through DataManager and shows it in standard UI components, the agent does not need to write permission checks for it.

Adaptation at runtime

Administrators and power users can make many changes in a running application, without new code or redeployment: create roles in the UI, filter data with the generic filter, track changes with audit, create reports, and add entities and attributes with the dynamic model.

Standard technology stack

A Jmix application is a Spring Boot application written in Java or Kotlin. The framework is open source, and the application code is yours. Your IDE, tests, static analysis and CI/CD pipelines work as usual. See Using Mainstream Technologies.

Developing with AI Agents

  • The AI Agent Setup section shows how to set up a project for an agent and how to work with the agent in it.

  • Jmix Agent Toolkit adds Jmix instructions, agent skills and MCP servers to your project. The AI Agents Toolkit action in Studio installs the toolkit in an existing project.

  • Jmix Docs MCP server lets the agent search the Jmix documentation, Java API and UI samples.

  • Jmix CLI creates projects from the command line. An agent can run it to create a project.

  • Jmix AI Assistant answers questions about Jmix in your browser and in Studio.

AI Features in Your Application

  • The AI Tools add-on adds a complete assistant based on Spring AI. It answers questions about the application data by calling tools, and it includes a chat UI.

  • The AI Chat add-on provides advanced UI components for chat interfaces. The application implements the connection to the model, so the components work with any LLM backend.

  • The AI-generated JPQL dataset type in the Reports add-on creates the query of a report band from a description in natural language.