1
Add it to Claude Code
claude mcp add -e "DB_URL=${DB_URL}" mcp-dummy-db -- python -m app.mainRequired:
DB_URL2
Make your agent remember this setup
mcp-dummy-db's config, env vars, and the gotchas you hit — recalled in every future Claude Code, Cursor, and Codex session.
npx conare@latestFree · one command · indexes the sessions already on disk. Set up in the browser instead →
What it does
- Secure connector layer between LLMs and PostgreSQL databases
- Prevents arbitrary SQL execution by using predefined tools
- Protects database credentials by keeping them in environment variables
- Uses parameterized queries to ensure safety against SQL injection
- Provides full logging and audit trails for database operations
Tools 3
get_employees_by_departmentRetrieves a list of employees belonging to a specific department.get_projects_by_statusRetrieves a list of projects filtered by their current status.get_issues_by_priorityRetrieves a list of issues filtered by their priority level.Environment Variables
DB_URLrequiredDatabase connection string for PostgreSQLTry it
→Fetch all employees currently working in the AI department.
→List all projects that have been marked as Completed.
→Show me all issues that are currently set to High priority.
Original README from muniasamyk/MCP_Project
MCP Agent POC
This project demonstrates a secure, production-ready implementation of the Model Context Protocol (MCP) as a connector layer between AI agents and PostgreSQL databases. The solution enables natural language queries without exposing database credentials to the LLM.
Key Achievement: LLM cannot access database directly - only through predefined MCP tools.
🏗️ Architecture Overview
┌────────────────────────────────────────────────────────┐
│ USER QUERY │
│ "Fetch employees in AI department" │
└───────────────────┬────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────┐
│ PLANNER AGENT (LLM) │
│ ✓ Natural Language Understanding │
│ ✗ NO database credentials │
│ Output: {"tool": "get_employees_by_department", │
│ "parameters": {"department": "AI"}} │
└───────────────────┬────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────┐
│ EXECUTOR AGENT │
│ ✓ Validates tool request │
│ ✓ Maps to allowed operations only │
│ ✗ Cannot execute arbitrary SQL │
└───────────────────┬────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────┐
│ MCP TOOLS LAYER (Sandbox) │
│ ✓ get_employees_by_department("AI") │
│ ✓ get_projects_by_status("Completed") │
│ ✓ get_issues_by_priority("High") │
│ ✗ Cannot run arbitrary SQL │
└───────────────────┬────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────┐
│ DATABASE CONNECTION (Secure) │
│ ✓ Credentials in environment variables │
│ ✓ Only parameterized queries (SQL injection safe) │
└───────────────────┬────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────┐
│ RESULT TO USER │
│ [Secure data retrieval via MCP] │
└────────────────────────────────────────────────────────┘
🔒 Security Features
| Feature | With MCP |
|---|---|
| DB Credentials | Secure in .env ✅ |
| SQL Access | Predefined tools only ✅ |
| Attack Surface | Limited operations only ✅ |
| Audit Trail | Full logging ✅ |
| Connection Pool | Yes ✅ |
Project Structure
app/agents/: The brain (Planner, Executor, Orchestrator)app/mcp/: The tool layer (Connector to DB)app/database/: Low-level DB connection poolapp/api/: FastAPI routes
Getting Started
1. Setup Env
Copy the example config:
cp .env.example .env
2. Run with Docker
The easiest way to stand it up (Postgres + API):
docker-compose up --build
The API listens on http://localhost:8000.
3. Test It
You can use the swagger UI at /docs or curl:
curl -X POST "http://localhost:8000/api/v1/query" \
-H "Content-Type: application/json" \
-d '{"query": "Find all projects that are in progress"}'
Local Dev (No Docker)
If you have Python 3.11+ and a local Postgres running:
pip install -r requirements.txt- Update
.envwith your DB credentials python -m app.main