Order_Food_MCP

Design principle: business logic (menu.py) is kept fully independent of MCP wiring (server.py), so it can be tested in milliseconds without a running server or LLM.

Menu Design

8 items, each with a unique id, name, and price (float, for now):

Related MCP server: MCPizza - Enhanced

Testing Strategy

Tests in test_menu.py cover:

• Menu size is within the 5–10 item range

• Every item has valid id/name/price

• All IDs are unique

• Lookup by valid ID returns the correct item

• Lookup by invalid ID returns None (not an exception)

Run with:

uv run pytest test_menu.py -v

Status

• [done] Week 1: MCP fundamentals, hello-world server, local LLM verified

• [done] Week 2: Menu data model, mock data, lookup function, tests passing

• [done] Week 3: get_menu MCP tool, tests, live LLM demo (Claude Desktop)

• Week 4: Order logic (pricing, validation)

Week 3 Demo

Verified get_menu_tool end-to-end in Claude Desktop: asked "What food can I order?" in natural language (no explicit tool mention), and the LLM correctly called get_menu_tool and returned the full 8-item menu in chat.

Also verified via MCP Inspector: tool discoverable, returns valid JSON with ( cmd - npx @modelcontextprotocol/inspector uv run server.py ) id/name/price for all 8 items.