Jmix AI-Assisted Development Template

A template project with pre-configured AI coding assistant rules for Jmix 2.7.x applications.

Prerequisites

• Java 17+
• Node.js 18+ (for MCP servers, optional)
• IntelliJ IDEA with Jmix Studio plugin

Quick Start

./gradlew bootRun     # http://localhost:8080 (admin/admin)
./gradlew test        # Run tests

Agentic AI Setup

Junie (Recommended)

Recommended for IntelliJ IDEA users due to native IDE and Jmix Studio integration.

Junie reads .junie/guidelines.md by default. To use the universal AGENTS.md instead:

Option A: Prompt prefix

Read AGENTS.md first, then proceed with the task. <Your prompt here for task>

Option B: Change default path

Settings → Tools → Junie → Project Settings → Guidelines path → set AGENTS.md (instead of .junie/guidelines.md)

Other Tools

Tool	Auto-reads	Notes
Claude Code	CLAUDE.md	Case-sensitive, only context file for vendor
Codex (OpenAI)	AGENTS.md	Works out of the box
Continue.dev	AGENTS.md, .continuerules	Also, works out of the box

If your tool uses a different config file and you don't want to always add pre-prompt about AGENTS.md, copy AGENTS.md content to that file (e.g., CLAUDE.md, .cursorrules) - or rename it to needing file.

MCP Setup (Optional)

MCP servers enable AI to verify code via IDE and browser. Requires Node.js 18+.

Step 1: Enable MCP Server in IntelliJ IDEA (required for JetBrains MCP)

Settings → Tools → MCP Server → Enable MCP Server ✓

Step 2: Configure MCP in your AI tool

Junie:

Settings → Tools → Junie → MCP Settings → Edit mcp.json:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["-y", "@playwright/mcp"]
    }
  }
}

Note: Junie runs inside IDEA, so JetBrains MCP is not needed.

Claude Code:

claude mcp add jetbrains -- npx -y @jetbrains/mcp-proxy
claude mcp add playwright -- npx -y @playwright/mcp

Codex (OpenAI):

codex mcp add jetbrains -- npx -y @jetbrains/mcp-proxy
codex mcp add playwright -- npx -y @playwright/mcp

Optional:

• Context7 MCP can be used for Jmix documentation search.

Step 3: Verify MCP is working

Ask your AI agent:

Do you have access to JetBrains MCP? Can you run get_file_problems()?
Do you have access to Playwright MCP? Can you take a screenshot?

If configured correctly, the agent will confirm available tools.

Example Workflow

You: Create a Customer entity with name and email fields, and list/detail views

AI: 1. Reads AGENTS.md (auto or via prompt prefix)
    2. Reads ai-docs/skills/entities/SKILL.md, ai-docs/skills/views/SKILL.md
    3. Creates Customer.java with @JmixEntity, UUID, @Version, @InstanceName
    4. Creates Liquibase changelog, adds to changelog.xml
    5. Creates CustomerListView.java + XML, CustomerDetailView.java + XML
    6. Updates menu.xml
    7. Adds messages to ALL locale files
    8. Validates against ai-docs/rules/entities/RULE.md, ai-docs/rules/views/RULE.md
    9. Runs get_file_problems() via JetBrains MCP (if configured)
   10. Runs tests
   11. Opens browser via Playwright MCP, navigates to view, verifies UI works (if configured)

After AI completes the task, you can ask for additional verification:

Check all files for IDE errors using get_file_problems()
Run the app and verify Customer views work via Playwright
Run all tests and show me the results

The AI reads the relevant skill (ai-docs/skills/{layer}/SKILL.md) and applies rules (ai-docs/rules/{layer}/RULE.md) based on the task type (layer = entities or views, services, etc.).

What's Inside

ai-docs/
├── skills/          # Step-by-step guides (entities, views, services...)
└── rules/           # Constraints applied by file path context

Skills — read when agent needs to perform a task (e.g., create entity) Rules — auto-applied based on file path (e.g., **/entity/** triggers entity rules)

See AGENTS.md for the complete AI context file.

Copy to Your Project

1. Copy AGENTS.md to your project root (or rename to your tool's config file)
2. Copy ai-docs/ folder