Superset MCP Server
# WIP: MCP Server Superset
基于 Apache Superset REST API 构建的 Model Context Protocol (MCP) 服务器端应用。
## 项目简介
这是一个基于 Apache Superset RESTAPI 在MCP上实现了通过大模型来让它进行基本的查询能力。
## 功能特性
- 查询数据库
- 查询表
- 查询字段
- 执行sql
## 环境要求
- Node.js >= 14.0.0
## 快速开始
### 直接使用
```bash
npx -y https://github.com/LiusCraft/superset-mcp-server
SUPERSET_URL
SUPERSET_USERNAME
SUPERSET_PASSWORD
鉴权方式：ladp
```
### 安装依赖
```bash
# 安装 Node.js 依赖
npm install
```
### 启动服务
```bash
# api client 测试
npm run src/examples/superset-example.ts
# 调试环境
npm run inspector
# 生产环境
npm run build
npm start
```
## 配置说明
项目配置文件位于 `config` 目录下，包括：
- 数据库配置
- API 配置
- 安全配置
## API 文档
参考superset官方 swagger文档
## 开发指南
### 目录结构
```
.
├── src/ # 源代码目录
├── src/examples # 封装的api客户端测试代码
├── src/services # 封装的api函数
├── src/utils # 封装的superset baseHttpClient
├── src/index.ts # 定义mcp接口
```
### 开发规范
- 遵循 ESLint 规范
- 使用 TypeScript 进行开发
- 遵循 Git Flow 工作流
## 部署
### mcp 部署
1. build the project
2. set mcp config:
```bash
node currentFolder/build/index.js
```
3. use the mcp
## 贡献指南
1. Fork 本仓库
2. 创建特性分支
3. 提交变更
4. 发起 Pull Request
## 许可证
[Apache License 2.0](LICENSE)
## 联系方式
如有问题，请提交 Issue 或联系项目维护者。
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose with no overlap: list-databases enumerates databases, list-tables shows tables within a database, list-fields details fields within a table, and query-superset executes queries. The hierarchical progression from databases to tables to fields to queries eliminates any ambiguity.
All tools follow a consistent verb_noun pattern using kebab-case (list-databases, list-fields, list-tables, query-superset). The verbs 'list' and 'query' are appropriately descriptive and applied consistently across the set, making the naming highly predictable.
With 4 tools, the count is reasonable for a Superset data exploration server, covering core operations like listing resources and querying. It feels slightly lean but not incomplete, as it supports basic workflows without unnecessary bloat. A few more tools (e.g., for metadata or schema operations) could enhance it, but it's well-scoped.
The tools provide a logical flow for data exploration: list databases, then tables, then fields, and finally query. Minor gaps exist, such as no explicit tools for creating or managing resources (e.g., dashboards or charts), but the core query and discovery operations are covered, allowing agents to navigate and query data effectively.