calculator-mcp-server
# Calculator MCP Server
A Model Context Protocol server exposing calculator operations,
built as a learning project.
## Tools
- add(a, b)
- subtract(a, b)
- multiply(a, b)
- divide(a, b)
- power(base, exponent)
- square_root(number)
- modulus(a, b)
## Run
uv run mcp dev src/calculator_mcp/server.py
## Test
uv run pytest tests/ -v
## Architecture
- `operations.py` — pure calculator logic, no MCP dependency
- `server.py` — FastMCP tool registration, thin wrappers over operations.py
- `exceptions.py` — custom exception hierarchy (CalculatorError base)
- `tests/` — pytest unit tests, independent of the MCP layer
## Connect to Claude Desktop
Add this to `claude_desktop_config.json`:
{
"mcpServers": {
"calculator": {
"command": "uv",
"args": ["run", "--directory", "/path/to/calculator-mcp-server", "mcp", "run", "src/calculator_mcp/server.py"]
}
}
}
This structure generalizes to any MCP server: keep domain logic
separate from protocol code, use typed exceptions per failure mode,
annotate tools honestly (readOnly/idempotent/destructive/openWorld),
and test the logic layer without needing a live server.TDQS
Scored across 7 tools
Each tool performs a distinct mathematical operation: add, subtract, multiply, divide, modulus, square_root, and power. There is no overlap or ambiguity between them, so an agent will never have to guess which tool to use for a given operation.
The names are clear and short, but not perfectly uniform: add, subtract, multiply, divide, and power are verbs, while modulus and square_root are nouns. However, this is a common and readable convention for math operations, and the intent is obvious.
With 7 tools, the server is well-scoped for a calculator. It covers basic arithmetic plus modulus, square root, and power, which is a reasonable and expected set for the purpose without unnecessary bloat.
The core arithmetic and common math operations are present, providing good coverage for a calculator server. Minor omissions like absolute value or factorial exist, but they are not essential for the primary purpose and agents can typically work around them.