Architecture Discovery MCP

English | 中文

English

Auto-discover enterprise architecture artifacts from Java/Maven source code. Scan your project and generate TOGAF-standard deliverables in seconds — application architecture, data architecture, and data lineage.

What It Does

Point it at a Java/Maven project and get:

Application Architecture (AA):

• AA-01 Application Module Inventory (Excel)

• AA-02 Function Item Inventory (Excel)

• AA-03 Sub-function Inventory (Excel)

• AA-04 Function Distribution (Excel)

• AA-05 Integration Inventory (Excel)

• AA-07 Application Architecture Diagram (PPTX)

• AA-08 Application Integration Diagram (PPTX)

Data Architecture (DA):

• DA-01 Conceptual Entity List (Excel)

• DA-02 Logical Entity List (Excel)

• DA-03 Physical Entity List (Excel)

• DA-04 Database Table List (Excel)

• DA-05 Data Source List (Excel)

• DA-06 Table-Function Relationship (Excel)

• DA-07 Data Dictionary (Excel)

• DA-CDM Conceptual Data Model Diagram (PPTX)

• DA-LDM Logical Data Model Diagram (PPTX)

• DA-Flow Data Flow Diagram (PPTX)

Data Lineage:

• DA-08 Field-Level Lineage Inventory (Excel)

• DA-09 Data Flow Inventory (Excel)

• DA-LINEAGE Data Lineage Diagram (PPTX)

• DA-IMPACT Impact Analysis Report (Excel, per field)

All generated from static analysis — no runtime required, no code leaves your machine.

Install

pip install architecture-discovery-mcp

Requires Python 3.11+.

Usage

As an MCP Server (recommended)

The engine runs as an MCP server over stdio. Configure it in your AI IDE:

Claude Code (~/.claude.json):

{
  "mcpServers": {
    "architecture-discovery": {
      "command": "architecture-discovery-mcp"
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "architecture-discovery": {
      "command": "architecture-discovery-mcp"
    }
  }
}

VS Code (.vscode/settings.json):

{
  "mcp": {
    "servers": {
      "architecture-discovery": {
        "command": "architecture-discovery-mcp"
      }
    }
  }
}

Then ask your AI assistant naturally:

"Scan D:/projects/my-java-app and generate the application architecture artifacts"

Available MCP Tools

All tools accept an optional locale parameter ("zh" or "en", default "zh").

As a Python Library

from pathlib import Path
from src.scanner.project_scanner import scan_project
from src.analyzers.module_analyzer import analyze_modules
from src.analyzers.api_analyzer import analyze_apis
from src.analyzers.integration_analyzer import analyze_integrations
from src.generators.app_architecture_generator import generate_app_architecture

project = scan_project(Path("your-java-project"))
project = analyze_modules(project)
endpoints = analyze_apis(project)
integrations = analyze_integrations(project)

out = Path("output")
generate_app_architecture(project, endpoints, integrations, out, locale="en")

Data Lineage

from src.generators.lineage_combined_generator import generate_data_lineage
from src.generators.da_impact_generator import generate_da_impact

project = scan_project(Path("your-java-project"))
project = analyze_modules(project)

result = generate_data_lineage(project, Path("output"), locale="en")
graph = result["graph"]

# Impact analysis for a specific field
impact_path = generate_da_impact(graph, "order-service.Order.id", Path("output"), locale="en")

Supported Projects

• Java/Maven projects (must have pom.xml)

• Spring Boot / Spring Cloud microservices

• Feign client integration detection

• JPA entity detection (@Entity, @Table, @Column)

• MyBatis mapper XML parsing (including <association> / <collection>)

• MapStruct @Mapping annotation lineage

• Setter/getter chain detection (BeanUtils.copyProperties)

• SQL field source analysis (INSERT...SELECT via sqlparse)

• MQ producer detection (KafkaTemplate, RabbitTemplate, @SendTo)

• @Scheduled ETL task detection

• Enum scanning and data dictionary generation

• Nested multi-module projects

Internationalization

Artifacts can be generated in Chinese (locale="zh", default) or English (locale="en"). This controls:

• Excel column headers and sheet names

• Cell values (layer names, status labels)

• PPTX slide titles and diagram labels

• Font selection (Microsoft YaHei for Chinese, Calibri for English)

Security

• Runs locally via stdio — no network exposure

• Source code never leaves your machine

• Exported JSON contains only structural metadata, not source code

Development

git clone https://github.com/rdtan/architecture-discovery-mcp.git
cd architecture-discovery-mcp
pip install -e ".[dev]"
pytest tests/ -v

License

MIT

Related MCP server: ArchiScribe MCP Server

中文

从 Java/Maven 源码自动发现企业架构制品。扫描项目即可在数秒内生成符合 TOGAF 标准的交付物——应用架构、数据架构和数据血缘。

功能概览

指向一个 Java/Maven 项目即可获得：

应用架构 (AA)：

• AA-01 应用模块清单 (Excel)

• AA-02 功能项清单 (Excel)

• AA-03 子功能清单 (Excel)

• AA-04 功能分布矩阵 (Excel)

• AA-05 集成清单 (Excel)

• AA-07 应用架构图 (PPTX)

• AA-08 应用集成图 (PPTX)

数据架构 (DA)：

• DA-01 概念实体清单 (Excel)

• DA-02 逻辑实体清单 (Excel)

• DA-03 物理实体清单 (Excel)

• DA-04 数据库表清单 (Excel)

• DA-05 数据来源清单 (Excel)

• DA-06 表-功能关系 (Excel)

• DA-07 数据字典 (Excel)

• DA-CDM 概念数据模型图 (PPTX)

• DA-LDM 逻辑数据模型图 (PPTX)

• DA-Flow 数据流转图 (PPTX)

数据血缘：

• DA-08 字段级映射清单 (Excel)

• DA-09 数据流转清单 (Excel)

• DA-LINEAGE 数据血缘图 (PPTX)

• DA-IMPACT 影响分析报告 (Excel，按字段)

全部基于静态分析生成——无需运行时环境，代码不会离开你的机器。

安装

pip install architecture-discovery-mcp

需要 Python 3.11+。

使用方式

作为 MCP Server（推荐）

本工具以 MCP 协议通过 stdio 运行。在 AI IDE 中配置：

Claude Code (~/.claude.json):

{
  "mcpServers": {
    "architecture-discovery": {
      "command": "architecture-discovery-mcp"
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "architecture-discovery": {
      "command": "architecture-discovery-mcp"
    }
  }
}

VS Code (.vscode/settings.json):

{
  "mcp": {
    "servers": {
      "architecture-discovery": {
        "command": "architecture-discovery-mcp"
      }
    }
  }
}

然后用自然语言对 AI 助手说：

"扫描 D:/projects/my-java-app 并生成应用架构制品"

可用 MCP 工具

所有工具支持可选参数 locale（"zh" 或 "en"，默认 "zh"）。

作为 Python 库使用

from pathlib import Path
from src.scanner.project_scanner import scan_project
from src.analyzers.module_analyzer import analyze_modules
from src.generators.lineage_combined_generator import generate_data_lineage
from src.generators.da_impact_generator import generate_da_impact

project = scan_project(Path("your-java-project"))
project = analyze_modules(project)

# 生成数据血缘制品
result = generate_data_lineage(project, Path("output"), locale="zh")
graph = result["graph"]

# 影响分析
impact_path = generate_da_impact(graph, "order-service.Order.id", Path("output"), locale="zh")

支持的项目类型

• Java/Maven 项目（需包含 pom.xml）

• Spring Boot / Spring Cloud 微服务

• Feign 客户端集成检测

• JPA 实体检测（@Entity、@Table、@Column）

• MyBatis Mapper XML 解析（含 <association> / <collection>）

• MapStruct @Mapping 注解血缘

• Setter/Getter 链检测（BeanUtils.copyProperties）

• SQL 字段来源分析（INSERT...SELECT，基于 sqlparse）

• MQ 生产者检测（KafkaTemplate、RabbitTemplate、@SendTo）

• @Scheduled ETL 任务检测

• 枚举扫描和数据字典生成

• 嵌套多模块项目

国际化

制品可以生成中文（locale="zh"，默认）或英文（locale="en"）版本，控制：

• Excel 列头和 Sheet 名

• 单元格内容（层级名称、状态标签）

• PPTX 幻灯片标题和图表标注

• 字体选择（中文用微软雅黑，英文用 Calibri）

安全性

• 本地通过 stdio 运行——无网络暴露

• 源代码不会离开你的机器

• 导出的 JSON 仅包含结构元数据，不含源代码

开发

git clone https://github.com/rdtan/architecture-discovery-mcp.git
cd architecture-discovery-mcp
pip install -e ".[dev]"
pytest tests/ -v

许可证

MIT