Setup & Installation
What This Skill Does
A development guide for building MCP (Model Context Protocol) servers that connect LLMs to external APIs and services. Covers the full lifecycle: planning tool design, implementing in TypeScript or Python, testing with MCP Inspector, and writing evaluations. Includes reference docs for both SDKs and a structured four-phase workflow.
Rather than reading scattered SDK READMEs and protocol specs separately, this guide consolidates tool design patterns, evaluation strategies, and language-specific examples into a single workflow that reduces the chance of building tools agents can't reliably use.
When to use it
- Building a TypeScript MCP server that exposes GitHub API endpoints to an LLM
- Designing tool schemas with Zod so agents receive structured, typed responses
- Writing 10-question evaluation suites to verify an MCP server works with real LLM prompts
- Choosing between stdio and streamable HTTP transport for local vs. remote deployments
- Debugging tool discoverability by reviewing naming conventions and description clarity