db-connector
Provides read-only access to PostgreSQL databases, enabling AI agents to list tables, inspect schemas, and run SELECT queries with result limits.
Provides read-only access to SQLite databases, enabling AI agents to list tables, inspect schemas, and run SELECT queries with result limits.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@db-connectorWhich city has the highest sales?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
db-connector · MCP 落地服务 Demo(TypeScript + DeepSeek/OpenRouter)
让 AI 直接查询你的业务数据库,全程只读、安全可控。 MCP 落地服务的敲门砖作品:TypeScript 实现 + AI 驱动,真实可跑。
技术栈
MCP 协议:官方
@modelcontextprotocol/sdk(server + client,标准协议)数据库:
better-sqlite3(演示)↔ PostgreSQL(生产,改适配层即可)AI:默认 DeepSeek,也支持 OpenRouter(含免费模型)——任意 OpenAI 兼容模型只需改
.env
Related MCP server: GraphJin
快速开始
npm install
cp .env.example .env # 填入 API key(DeepSeek 或 OpenRouter,见下)
npm run seed # 生成演示库 demo.db(电商订单系统)
npm run smoke # 冒烟测试(不依赖 key,验证协议与只读防护)
npm run e2e # 端到端测试(3 个自然语言问题,需 key)
npm run chat # 命令行直接问数据库(AI 自动查库回答)> 哪个城市销售额最高?
🤖 销售额最高的城市是南京,约 ¥8.7 万(59 单)……
> 最近 30 天卖得最好的商品是?
> 买得最多的 3 个客户?配置模型(三选一)
# ① DeepSeek 官方(推荐:便宜、快、国内可用)
DEEPSEEK_API_KEY=sk-xxx
DEEPSEEK_BASE_URL=https://api.deepseek.com
DEEPSEEK_MODEL=deepseek-chat
# ② OpenRouter 免费模型(demo 零成本,但每天限 50 次请求,充值 $10 解锁 1000 次)
DEEPSEEK_API_KEY=sk-or-v1-xxx
DEEPSEEK_BASE_URL=https://openrouter.ai/api/v1
DEEPSEEK_MODEL=nvidia/nemotron-3-super-120b-a12b:free # 实测工具调用较稳
# ③ 本地 ollama(零成本、数据不出本机,需先装 ollama 并 pull qwen3:8b)
DEEPSEEK_API_KEY=
DEEPSEEK_BASE_URL=http://localhost:11434/v1
DEEPSEEK_MODEL=qwen3:8b
# ④ 任意 OpenAI 兼容服务(通义/豆包/GPT 等)
DEEPSEEK_BASE_URL=<你的服务地址>
DEEPSEEK_MODEL=<模型名>本地模型演示建议:qwen3:8b(
ollama pull qwen3:8b)经实测能稳定答对「哪个城市销售额最高」「哪个商品卖得最多」,但涉及日期函数/复杂多表的问题会不稳(8B 能力边界)。演示脚本请用稳定问题集。
两种演示方式(给客户看)
方式一:CLI 直接问(推荐,最快出效果)
npm run chat → 输入问题 → AI 自动完成「找表 → 看结构 → 写 SQL → 查库 → 分析回答」。全程肉眼可见地调用工具。
方式二:接入 Claude Desktop(展示 MCP 通用性)
编辑 ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"db-connector": {
"command": "/Users/wjy/.workbuddy/binaries/node/versions/22.22.2/bin/node",
"args": ["--import", "tsx", "/Users/wjy/Documents/code/apps/mcp-demo-server/src/server.ts"]
}
}
}重启后直接问:「帮我分析订单数据,哪个城市销售额最高?」(推荐先用 npm run build 编译成 dist 再用 node 跑,避免依赖 tsx)
演示问题清单
「按城市统计销售额排名,哪个城市卖得最好?」
「最近 30 天销售额和订单量是多少?环比上月增长多少?」
「哪个商品是爆款?哪个商品卖不动?」
「帮我查出买得最多的 3 个客户」
架构
AI 客户端 (DeepSeek Agent / Claude Desktop / Cursor)
│ MCP 标准协议
▼
db-connector (本仓库)
├── list_tables → 自动发现数据结构
├── describe_table → 查看表结构
└── query → 只读 SQL 查询(安全拦截非 SELECT)
│
▼
SQLite (demo.db) / PostgreSQL (生产)to B 安全卖点(客户最关心):
能力 | 实现 |
只读防护 | 非 |
结果限额 | 默认 20 行、最多 100 行,防拖库 |
模型自由 | DeepSeek / 任意 OpenAI 兼容模型,不绑定 Claude 订阅 |
协议标准 | MCP 官方协议,所有 AI 客户端通用 |
生产扩展(接单后按客户需求加)
认证与权限:按客户/角色限制可访问的表(白名单)
操作日志:记录 AI 执行的每条 SQL(审计)
只读账号:接数据库只读副本,从源头杜绝写操作
更多工具:除数据库外可接 ERP/CRM API、内部知识库、审批流
目录结构
src/
├── db.ts # 数据库适配 + 只读防护
├── server.ts # MCP Server(stdio 模式 + setupServer 复用)
├── agent.ts # DeepSeek Agent(自然语言 → MCP 工具 → 回答)
├── cli.ts # 命令行问答入口
├── seed.ts # 生成演示数据库
└── smoke.ts # 冒烟测试(不依赖 API key)常见问题
Q: AI 会不会乱改数据? 不会。服务端强制只读:非查询类 SQL 直接拦截报错;生产再加只读账号双保险。
Q: 必须用 DeepSeek 吗? 不用。改 .env 里的 DEEPSEEK_BASE_URL / DEEPSEEK_MODEL 就能换任意 OpenAI 兼容模型(通义/豆包/GPT 等)。
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- FlicenseAqualityDmaintenanceEnables AI assistants to interact with PostgreSQL databases using natural language queries, providing secure read-only access to database schemas and SQL translation capabilities.610
- Alicense-qualityAmaintenanceEnables AI assistants to query databases using natural language, with automatic schema discovery and SQL compilation.6043,135Apache 2.0
- Alicense-qualityDmaintenanceEnables AI assistants to connect to and interact with PostgreSQL, MySQL, SQLite, and MongoDB databases through natural language, supporting schema exploration, query execution, data export, and more.MIT
- Alicense-qualityBmaintenanceEnables AI agents to securely query databases (PostgreSQL, SQLite, MySQL, DuckDB) with read-only defaults and multi-layer SQL injection prevention.1MIT
Related MCP Connectors
Query PostgreSQL databases in plain English — LLM-generated, safety-validated SQL.
The grounded data layer for any LLM: governed SQL, metrics, lineage and catalog over your data.
Analytical memory for AI agents: a real Postgres queried in plain English over MCP. One command.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/adlkt/mcp-demo-server'
If you have feedback or need assistance with the MCP directory API, please join our Discord server