Oracle Database MCP Server

Oracle Database MCP Server

Enables GitHub Copilot and other LLMs to execute read-only SQL queries against Oracle databases with secure connection pooling and schema introspection capabilities.

Category
访问服务器

README

Database MCP Server

A Model Context Protocol (MCP) server that enables GitHub Copilot and other LLMs to execute read-only SQL queries against Oracle databases.

npm version License: ISC

📦 Installation

From npm (Recommended)

npm install -g mcp-oracle-database

Or install locally in your project:

npm install mcp-oracle-database

From Source

git clone https://github.com/tannerpace/my-mcp.git
cd my-mcp
npm install && npm run build

�🚀 Quick Start with VS Code

If installed via npm:

  1. Configure VS Code MCP settings

Create .vscode/mcp.json:

{
  "servers": {
    "oracleDatabase": {
      "type": "stdio",
      "command": "mcp-database-server",
      "env": {
        "ORACLE_CONNECTION_STRING": "localhost:1521/XEPDB1",
        "ORACLE_USER": "your_readonly_user",
        "ORACLE_PASSWORD": "your_password",
        "ORACLE_POOL_MIN": "2",
        "ORACLE_POOL_MAX": "10",
        "QUERY_TIMEOUT_MS": "30000",
        "MAX_ROWS_PER_QUERY": "1000"
      }
    }
  }
}
  1. Reload VS Code and ask Copilot:
"What tables are in the database?"

If running from source:

# 1. Build the server
npm install && npm run build

# 2. Configure VS Code
cp .vscode/mcp.json.example .vscode/mcp.json

# 3. Start Oracle database (if using Docker)
docker start oracle-xe

# 4. Reload VS Code and ask Copilot:
"What tables are in the database?"

See Quick Start Guide for detailed setup.

Features

  • 🔒 Read-only access - Uses a dedicated read-only database user for security
  • 📡 stdio transport - Communicates via standard input/output (no HTTP server needed)
  • Connection pooling - Efficient Oracle connection management
  • 📊 Schema introspection - Query table and column information
  • 📝 Audit logging - All queries are logged with execution metrics
  • ⏱️ Timeout protection - Prevents long-running queries
  • 🛡️ Result limits - Configurable row limits to prevent memory issues

Architecture

GitHub Copilot
      ↓ (MCP Protocol)
  MCP Client (spawns process)
      ↓ (JSON-RPC over stdio)
  MCP Server (Node.js)
      ↓ (oracledb)
  Oracle DB (read-only user)

Prerequisites

  1. Node.js v18 or higher
  2. Oracle Database with a read-only user created
    • Running locally (Docker recommended)
    • Or accessible remote instance

Note: This project uses the node-oracledb package in Thin Mode, which means no Oracle Instant Client installation is required! The pure JavaScript driver connects directly to Oracle Database, just like Python's oracledb library.

Optional: Running Oracle Database Locally with Docker

If you need a local Oracle database for development:

macOS (using Colima):

# Start Colima (Docker runtime for macOS)
colima start

# Pull and run Oracle XE container
docker run -d \
  --name oracle-xe \
  -p 1521:1521 \
  -p 5500:5500 \
  -e ORACLE_PWD=OraclePwd123 \
  container-registry.oracle.com/database/express:latest

# Wait for database to be ready (takes 1-2 minutes)
docker logs -f oracle-xe

# Start/stop the database later
docker start oracle-xe
docker stop oracle-xe

Linux/Other:

# Same docker commands as above, just ensure Docker is running
docker ps

The database will be available at:

  • Connection: localhost:1521/XEPDB1
  • SYS password: OraclePwd123
  • Web UI: http://localhost:5500/em

Setup

1. Clone and Install Dependencies

git clone <repository-url>
cd my-mcp
npm install

2. Create Read-Only Database User

Connect to your Oracle database as a DBA and run:

-- Create read-only user
CREATE USER readonly_user IDENTIFIED BY secure_password;

-- Grant connect and read-only privileges
GRANT CONNECT TO readonly_user;
GRANT SELECT ANY TABLE TO readonly_user;

-- Or grant access to specific tables only:
GRANT SELECT ON schema.table1 TO readonly_user;
GRANT SELECT ON schema.table2 TO readonly_user;

3. Configure Environment Variables

Copy the example environment file:

cp .env.example .env

Edit .env and set your Oracle connection details:

# Oracle Database Connection (READ-ONLY USER)
ORACLE_CONNECTION_STRING=hostname:1521/servicename
ORACLE_USER=readonly_user
ORACLE_PASSWORD=secure_password

# Connection Pool Settings
ORACLE_POOL_MIN=2
ORACLE_POOL_MAX=10

# Query Settings
QUERY_TIMEOUT_MS=30000
MAX_ROWS_PER_QUERY=1000
MAX_QUERY_LENGTH=50000

# Logging
LOG_LEVEL=info
ENABLE_AUDIT_LOGGING=true

4. Build the Server

npm run build

5. Configure GitHub Copilot / MCP Client

Create or update your MCP client configuration file:

VS Code (cline_mcp_config.json or similar):

{
  "mcpServers": {
    "oracle-db": {
      "command": "node",
      "args": ["/absolute/path/to/my-mcp/dist/server.js"],
      "env": {
        "ORACLE_CONNECTION_STRING": "hostname:1521/servicename",
        "ORACLE_USER": "readonly_user",
        "ORACLE_PASSWORD": "secure_password"
      }
    }
  }
}

Or use environment variables from your shell:

{
  "mcpServers": {
    "oracle-db": {
      "command": "node",
      "args": ["/absolute/path/to/my-mcp/dist/server.js"]
    }
  }
}

Usage

Once configured, the MCP server provides two tools to GitHub Copilot:

Testing with the Built-in Client

Before integrating with Copilot, you can test the server locally:

# Make sure you have .env configured with valid Oracle credentials
npm run build
npm run test-client

This will:

  1. Start the MCP server
  2. Connect to it via the test client
  3. List available tools
  4. Get database schema (list all tables)
  5. Disconnect and shut down

Edit src/client.ts to customize the test queries.

Using with GitHub Copilot

Once configured, the MCP server provides two tools to GitHub Copilot:

1. query_database

Execute read-only SQL queries:

User: "Show me the top 10 customers by revenue"
Copilot: [calls query_database with SQL query]

Parameters:

  • query (required): SQL SELECT statement
  • maxRows (optional): Maximum rows to return
  • timeout (optional): Query timeout in milliseconds

Example:

{
  "query": "SELECT customer_name, SUM(revenue) as total FROM customers GROUP BY customer_name ORDER BY total DESC",
  "maxRows": 10
}

2. get_database_schema

Get schema information:

User: "What columns are in the CUSTOMERS table?"
Copilot: [calls get_database_schema with tableName="CUSTOMERS"]

Parameters:

  • tableName (optional): Specific table name, or omit to list all tables

Example Prompts for Copilot

  • "List all tables in the database"
  • "Show me the schema of the ORDERS table"
  • "How many active users do we have?"
  • "What are the top 5 products by sales this month?"
  • "Show me recent transactions for customer ID 12345"

Development

Project Structure

my-mcp/
├── src/
│   ├── server.ts              # Main MCP server entry point
│   ├── client.ts              # Test client for local testing
│   ├── config.ts              # Configuration with Zod validation
│   ├── database/
│   │   ├── types.ts           # TypeScript types
│   │   ├── oracleConnection.ts # Connection pool manager
│   │   └── queryExecutor.ts   # Query execution logic
│   ├── tools/
│   │   ├── queryDatabase.ts   # query_database tool
│   │   └── getSchema.ts       # get_database_schema tool
│   └── logging/
│       └── logger.ts          # Winston-based logging
├── dist/                      # Compiled JavaScript (generated)
├── .env                       # Environment variables (git ignored)
├── .env.example               # Environment template
└── package.json

Scripts

npm run build       # Compile TypeScript
npm run dev         # Watch mode compilation
npm run clean       # Remove dist folder
npm run typecheck   # Type check without compiling
npm start           # Run the server (after building)
npm run test-client # Run test client to verify server works

Logging

All queries and events are logged in JSON format. Logs go to stdout/stderr:

{
  "level": "info",
  "message": "Query executed successfully",
  "timestamp": "2025-10-24T10:30:00.000Z",
  "audit": true,
  "query": "SELECT * FROM customers WHERE...",
  "rowCount": 42,
  "executionTime": 156
}

Set LOG_LEVEL=debug in .env for more verbose logging.

Security Considerations

  1. Read-Only User - Database user has only SELECT privileges
  2. Local Client - Designed for trusted local use only
  3. No Injection Protection - Trust the LLM to generate valid queries
  4. Query Limits - Row count and timeout limits prevent resource exhaustion
  5. Audit Logging - All queries logged for review

Troubleshooting

Docker/Colima Issues (macOS)

Docker not running:

# Check if Colima is running
colima status

# Start Colima if needed
colima start

# Verify Docker works
docker ps

Database won't start:

# Check container status
docker ps -a | grep oracle

# View logs
docker logs oracle-xe

# Restart if needed
docker restart oracle-xe

Connection Failed

Error: ORA-12545: Connect failed because target host or object does not exist

Solutions:

  • Check your ORACLE_CONNECTION_STRING format: hostname:port/servicename
  • For local Docker: use localhost:1521/XEPDB1
  • Verify database is running: docker ps | grep oracle

Permission Denied

Error: ORA-00942: table or view does not exist

Solution: Grant SELECT privileges to your read-only user on the required tables.

Database Not Ready

If the test client fails immediately after starting the database:

  • Wait 1-2 minutes for Oracle to fully initialize
  • Check health status: docker ps should show (healthy)
  • Watch startup logs: docker logs -f oracle-xe

Thin Mode vs Thick Mode

This project uses Thin Mode (pure JavaScript, no Oracle Client needed). If you encounter issues and want to use Thick Mode:

  1. Install Oracle Instant Client
  2. Add to your code: oracledb.initOracleClient() before creating the pool

For most use cases, Thin Mode is simpler and works great!

Documentation

📚 Integration Guides:

📊 Test Results:

  • Test Results - Comprehensive test results with example queries

📝 Custom Instructions:

License

ISC

Contributing

Contributions welcome! Please open an issue or pull request.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选