mcp-builder

Guide creation of MCP servers in TypeScript or Python with tool schemas and evaluations.

Updated May 15, 2026
One-click install
npx skills add https://github.com/lth0/codexSkill --skill mcp-builder-lth0
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-builder
Source: https://github.com/lth0/codexSkill/tree/main/skills/agents/mcp-builder
Command: npx skills add https://github.com/lth0/codexSkill --skill mcp-builder-lth0

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill helps you design and implement high-quality MCP (Model Context Protocol) servers so LLM agents can reliably call well-structured tools to accomplish real tasks.

Core Features & Use Cases

  • Tool-first MCP server design: Plans an API coverage strategy versus workflow tools to maximize agent effectiveness and composability.
  • Protocol- and SDK-aligned implementation guidance: Recommends practical architecture choices for TypeScript/Node or Python/FastMCP, including transports like streamable HTTP or stdio.
  • Production-ready tool contracts: Focuses on input/output schemas, naming conventions, pagination, actionable error messages, and tool annotations (readOnly/destructive/idempotent hints).
  • Evaluation creation: Guides creation of stable, read-only evaluation questions (XML format) to test whether agents can use the MCP server to answer realistic questions.

Quick Start

Use this skill to produce a complete MCP server plan and tool specification for a new external API integration by asking it to draft the Phase 1 planning checklist and the tool schemas for the first 10 endpoints you want your agents to use.

Frequently Asked Questions about mcp-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build an MCP server with TypeScript for external API integration?

Building an MCP server with TypeScript involves using the MCP TypeScript SDK to design well-structured tools, input/output schemas, and actionable error handling. You select transports like stdio or streamable HTTP to let LLM agents reliably call external APIs.

What is the best way to design tool schemas for an MCP server?

The best way to design MCP tool schemas is to follow protocol conventions for naming and descriptions, implement pagination, and include tool annotations like readOnly or destructive. This approach maximizes agent effectiveness and tool composability.

Can I use FastMCP to create MCP servers in Python?

Yes, you can use FastMCP to build MCP servers in Python. The workflow guides practical architecture choices for Python, ensuring your tools have production-ready contracts, pagination support, and actionable error messages for agents.

How do I generate evaluations for an MCP server?

You generate MCP server evaluations by creating stable, read-only questions in XML format. These evaluation tasks test whether LLM agents can successfully use your MCP server tools to answer realistic questions and accomplish actual tasks.

When should I use streamable HTTP instead of stdio for MCP transport?

You should choose streamable HTTP or stdio for MCP transport based on your deployment architecture needs. The protocol-aligned implementation guidance recommends practical transport selections to ensure your server communicates effectively with agents.