oradb-explorer-mcp
Enables read-only exploration of Oracle databases through natural language, providing schema inspection and safe bounded SQL query execution.
README
oradb-explorer-mcp
An MCP server and skills package for read-only Oracle exploration using
node-oracledb 7 in Thin mode.
Download the repository
Windows PowerShell:
git clone https://github.com/YOUR_USER/oradb-explorer-mcp.git
Set-Location oradb-explorer-mcp
macOS/Linux:
git clone https://github.com/YOUR_USER/oradb-explorer-mcp.git
cd oradb-explorer-mcp
Setup
Windows PowerShell:
npm install
Copy-Item .env.example .env
macOS/Linux:
npm install
cp .env.example .env
Windows .env:
ORACLE_USER=your_user
ORACLE_PASSWORD=your_password
ORACLE_DSN=your_connect_string
# Optional: used only if the normal Thin-mode connection fails
ORACLE_WALLET_LOCATION=C:\secure\adb-wallet
macOS/Linux .env:
ORACLE_USER=your_user
ORACLE_PASSWORD=your_password
ORACLE_DSN=your_connect_string
# Optional: used only if the normal Thin-mode connection fails
ORACLE_WALLET_LOCATION=/home/your-user/adb-wallet
Never commit the real .env file.
Start locally
Windows and macOS/Linux:
npx tsx mcp/server.ts
The server uses MCP stdio and normally produces no terminal output.
Codex MCP configuration
Windows:
[mcp_servers.oradb]
command = 'C:\path\to\oradb-explorer-mcp\node_modules\.bin\tsx.cmd'
args = ['C:\path\to\oradb-explorer-mcp\mcp\server.ts']
env = { ORACLE_ENV_FILE = '.env' }
startup_timeout_sec = 30
tool_timeout_sec = 120
macOS/Linux:
[mcp_servers.oradb]
command = '/path/to/oradb-explorer-mcp/node_modules/.bin/tsx'
args = ['/path/to/oradb-explorer-mcp/mcp/server.ts']
env = { ORACLE_ENV_FILE = '.env' }
startup_timeout_sec = 30
tool_timeout_sec = 120
Available tools
oradb_connect
oradb_status
oradb_disconnect
oradb_tables
oradb_describe
oradb_indexes
oradb_sql_preview
oradb_run_readonly_sql
oradb_dataframe_preview
oradb_dataframe_export
Natural-language workflow
The MCP server does not use Select AI or generate SQL inside the database. The host agent interprets the user’s natural-language request and generates a candidate SQL statement. The MCP server validates and executes it safely.
Recommended workflow:
1. User describes the desired analysis.
2. The agent generates one read-only SELECT or WITH statement.
3. The agent calls oradb_sql_preview.
4. The SQL and bind values are reviewed.
5. The agent calls oradb_run_readonly_sql with a row limit.
Example request:
Find the five most recent active customers and return their ID, name,
status, and registration date.
The agent may generate:
SELECT customer_id, name, status, registration_date
FROM customers
WHERE status = :status
ORDER BY registration_date DESC
FETCH FIRST 5 ROWS ONLY
with binds:
{
"status": "ACTIVE"
}
Schema exploration examples
List tables whose names start with CUSTOM.
Describe the CUSTOMERS table.
List indexes on APP_SCHEMA.CUSTOMERS.
Dataframe examples
Preview selected columns:
Preview CUSTOMER_ID, NAME, and STATUS from CUSTOMERS, limited to 20 rows.
Export a bounded result to a user-selected file:
Export CUSTOMER_ID, NAME, and STATUS from CUSTOMERS to
/tmp/customers.csv using oradb_dataframe_export.
Windows example:
Export CUSTOMER_ID, NAME, and STATUS from CUSTOMERS to
C:\\temp\\customers.csv using oradb_dataframe_export.
Supported formats are CSV and JSON. The result can then be consumed by pandas, Polars, or marimo without overwhelming the agent context.
Safety boundary
The server allows one bounded SELECT or WITH statement. It rejects
semicolon-separated statements and does not provide DDL, DML, PL/SQL, commit,
privilege, or credential-management tools.
Credentials, wallet files, database dumps, and generated artifacts must remain outside Git.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。