test-emp-database-mcp

An employee-analytics MCP server (HTTP transport) backed by SQLite. Unlike a generic SQL toolkit, every tool here runs a specific, hand-written query demonstrating GROUP BY, window functions, NULL handling, and subqueries.

The database self-seeds 13 sample employees across Engineering, Sales, Marketing, and HR (three of them interns with NULL salary), so the analytics return meaningful results immediately.

Schema

employees(id, name, department, salary) — salary is nullable; interns are NULL.

Related MCP server: DB MCP (HR CSV to SQLite)

Tools

Basic: add_employee, list_employees

Analytics:

• top_n_by_salary(n, department) — top N earners overall or within a department

• average_salary_by_department — GROUP BY, AVG (ignores interns), headcount

• department_salary_stats — count / min / max / avg per department

• salary_rank_within_department — window RANK() OVER (PARTITION BY ...)

• highest_paid_per_department — window ROW_NUMBER() = 1

• employees_above_company_average — above the true company-wide average

• salary_share_within_department — window SUM() OVER (PARTITION BY ...), % of dept payroll

• list_interns — employees with NULL salary

• headcount_by_department — COUNT per department

Run / test locally

cd ~/Documents/Claude/test-emp-database-mcp
uv add fastmcp
uv run main.py          # serves http://127.0.0.1:8000/mcp (leave running)

To test the tools, point the MCP Inspector at the running URL:

npx @modelcontextprotocol/inspector
# transport: Streamable HTTP, URL: http://127.0.0.1:8000/mcp

Deploy to FastMCP Cloud

1. Lock deps so fastmcp is in the lockfile: uv lock

2. Push main.py, pyproject.toml, uv.lock to a GitHub repo.

3. On fastmcp.cloud, point at the repo with entrypoint main.py.

Notes:

• The mcp object name and from fastmcp import FastMCP import are what FastMCP Cloud requires.

• The platform serves the mcp object itself; the __main__ block (local host/port) is ignored in the cloud.

• emp.db is git-ignored and re-seeded on startup.