# JeecgBoot

> Java low-code platform with a built-in LLM application layer

- Category: App UIs & Builders / Low-Code & No-Code Builders
- Language: Java
- Licence: Apache-2.0
- Repository: https://github.com/jeecgboot/JeecgBoot
- Website: https://jeecg.com
- Canonical page: https://ai-tldr.dev/tools/jeecgboot/

## In short

JeecgBoot is a Java-first low-code platform that has grown an AI layer. It sits in the Low-Code & No-Code Builders category of the App UIs & Builders tools and is open-source, written in Java and licensed Apache-2.0. To get started: clone the repository.

## Overview

JeecgBoot is a Java-first low-code platform that has grown an AI layer. The original product is a conventional enterprise stack — Spring Boot with MyBatis-Plus on the back end, Vue 3 with TypeScript, Vite and Ant Design Vue on the front — wrapped around an online designer for forms, reports, dashboards and Flowable workflows, and a code generator that turns a table definition into working CRUD screens. Its stated development model is deliberately hybrid: generate, configure online, then merge hand-written code, so you are not locked out of the source.

The AI half is a separate application platform bundled into the same product, comparable to Dify in shape: an LLM app builder with chat assistants, knowledge-base Q&A over RAG, flow orchestration, model management, vector-store connections and MCP plugin support. It is built on langchain4j, and the README lists ChatGPT, DeepSeek, Qwen, Zhipu, MiniMax and local models as supported backends. The two halves meet in features like AI table creation, AI-generated reports and natural-language form design.

Deployment covers both single-application and microservice modes — Spring Cloud Alibaba with Nacos, Gateway, Sentinel and SkyWalking — and the project ships Docker Compose scripts for each so switching between them is a configuration decision rather than a rewrite. It is broad rather than narrow: multi-tenant SaaS support, RBAC down to row, column and form-field level, multi-database compatibility across MySQL, PostgreSQL, Oracle, SQL Server, MariaDB, DM and KingbaseES, and a companion UniApp mobile framework. Apache 2.0 licensed.

## What it does

- Online designers for forms, reports, dashboards and Flowable workflows, with no code required for the common cases
- Code generator producing single-table, tree, one-to-many and one-to-one CRUD from a table definition
- Built-in LLM application platform: chat assistants, RAG knowledge base, flow orchestration, model and vector-store management, MCP plugins
- langchain4j backend supporting ChatGPT, DeepSeek, Qwen, Zhipu, MiniMax and local models
- AI-assisted table creation, report generation and natural-language form design
- Single-application or Spring Cloud Alibaba microservice mode from the same codebase
- Fine-grained RBAC to row, column and form-field level, plus multi-tenant SaaS support

## Getting started

Two supported paths from the official docs: a Docker Compose script for a running system in one step, or a manual build if you want to develop against it. Building by hand needs JDK 17 or newer, Maven, Node 20+ and pnpm 9+.

1. Clone the repository

```bash
git clone https://github.com/jeecgboot/JeecgBoot.git
cd JeecgBoot
```

2. One-click Docker start (monolith)

   Run the script for your platform: `start-docker-compose.sh` on Linux, `start-docker-compose.bat` on Windows. The microservice equivalents are `start-docker-compose-cloud.sh` / `.bat`.

```bash
./start-docker-compose.sh
```

3. Or build back end and front end yourself

   The `-Pdocker` profile packages the backend for the compose stack; the frontend build writes the assets it serves.

```bash
cd JeecgBoot/jeecg-boot
mvn clean install -Pdocker
cd ../jeecgboot-vue3
pnpm i
pnpm run build:docker
cd ..
docker-compose up -d
```

4. Wait for MySQL, then open it

   The docs call this out explicitly: give MySQL about a minute to finish initialising its data before you load the page, or the first request fails.

```text
# http://localhost:80
```

5. Connect a model to the AI platform

   Model providers are configured from the AI application platform's model management screen — ChatGPT, DeepSeek, Qwen, Zhipu, MiniMax or a local endpoint. The AI handbook at https://help.jeecg.com/aigc walks through the knowledge base and app builder.

## When to use it

- Stand up an internal Java system — forms, reports, workflow, permissions — without writing the CRUD layer
- Add a RAG knowledge base and chat assistant to an existing enterprise app rather than running Dify beside it
- Generate a screen from a table, then hand-edit the generated code when the requirements outgrow the designer
- Start as a single application and move to Spring Cloud Alibaba microservices without changing codebases