Build Your First MCP Server

PROJECT CONTEXT:
I want my agent's tools available over MCP so any compatible host can use them. The MCP ecosystem shows the shape: a server declaring typed tools, a transport, and hosts that discover and call them.

GOAL:
Build an MCP server with 2-3 real tools, connect it to a host/client, and verify discovery, invocation, and error behavior end to end.

REQUIREMENTS:
- An MCP server exposing 2-3 genuinely useful tools with typed schemas
- Correct protocol lifecycle: initialization, tool listing, tool calls
- Meaningful tool results and structured errors over the protocol
- A client/host connection proving discovery and invocation
- Server-side logging of every call for debugging

CONSTRAINTS:
- Use the official SDK for your language rather than hand-rolling the protocol
- Tools must validate inputs server-side - never trust the client's model

DEFINITION OF DONE:
- The host lists the server's tools with correct names, descriptions, and schemas
- Each tool executes and returns correct results through the protocol
- Invalid arguments produce protocol-level errors the model can read, not crashes
- The call log shows arguments and results for every invocation

COMMON FAILURES TO AVOID:
- Hand-rolled protocol handling that breaks on the first spec nuance
- Tools that work in isolation but fail through the host's serialization
- Errors thrown as exceptions that kill the server instead of error results
- No logging, making host-side failures undiagnosable