# Oracle Database

Oracle Database — an MCP server that puts Oracle behind your agent: bound SQL queries, table and index lookup across schemas, plus Lab control-plane tools.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: samscarrow
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/samscarrow/oracle-mcp-server
- Updated: 2026-10-29
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/oracle-database

## What it does

\`execute_query` runs SQL against Oracle and accepts bind parameters; Postgres-style `$1` placeholders get rewritten to `:1` before the call goes out. The rest of the generic set reads the catalog: `list_tables` across schemas, `describe_table` for columns and defaults, `list_schemas`, `get_table_indexes`, `get_table_constraints`. A second group talks to the Lab control-plane tables — `check_gates`, `get_dispatchable_items`, `build_dispatch_packet`, `stamp_dispatch_consumed`, `fail_dispatch_preflight`, `handle_final_return`, `dispatch_scene`.

Setup starts with `npm install` in the repo. Credentials live in `.env`, and there are three ways to supply them: an Easy Connect string, a TNS name, or host, port and service name as separate variables. `ORACLE_DEFAULT_SCHEMA` picks the schema when it differs from the user. Register the server in `~/.claude.json` under `mcpServers` with `"type": "stdio"` and a `node` command pointing at `src/index.js`, then restart Claude. The Lab mirror comes from the DDL files, run through `sqlplus` on Oracle or `psql` on PostgreSQL.

You need Node.js 18+, an Oracle instance and a route to it. Connections open per query rather than from a pool. `execute_query` also lets writes through: bind variables block injection, they do not stop a stray `DELETE`, and the audit log only records timing and duration.

## Requirements
- Needs: Docker

## Install

### Claude Code (native platform)

**Install**
- The MCP registry has no launch command for it: see the repository README

## When to use it
- inspect a table's columns without opening SQL*Plus
- list tables in one schema with row counts
- run a parameterised query instead of concatenating values
- trace indexes and foreign keys on a table
- read `lab_work_items` before starting a dispatch run

## Example prompts
- List every table in the `HR` schema with its row count.
- Describe `EMPLOYEES`: columns, types, defaults.
- Run a query against `ORDERS` with `:1` as a bind parameter.

## Similar
- [oracle-dba](https://affpapa.org/en/nevibecode/skills/oracle-dba.md) — skill · 145 installs
- [oracle-ai-data-platform-workbench-engineer-agent](https://affpapa.org/en/nevibecode/plugins/oracle-ai-data-platform-workbench-engineer-agent.md) — plugin · ★ 62 repo stars
- [reviewing-oracle-to-postgres-migration](https://affpapa.org/en/nevibecode/skills/reviewing-oracle-to-postgres-migration.md) — skill · 1800 installs
- [Devart MCP Server for Oracle (ADO.NET)](https://affpapa.org/en/nevibecode/mcp/devart-mcp-server-for-oracle-adonet.md) — MCP server
- [database-management](https://affpapa.org/en/nevibecode/skills/database-management.md) — skill · 108 installs
- [database](https://affpapa.org/en/nevibecode/mcp/database.md) — MCP server · ★ 10 repo stars

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
