mcp-builder

Build MCP servers in Python or TypeScript with schemas and evaluation harnesses.

Updated Mar 31, 2026
One-click install
npx skills add https://github.com/DFly7/iOS-FastAPI-Supabase-AI --skill mcp-builder-dfly7
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-builder
Source: https://github.com/DFly7/iOS-FastAPI-Supabase-AI/tree/main/.agents/skills/mcp-builder
Command: npx skills add https://github.com/DFly7/iOS-FastAPI-Supabase-AI --skill mcp-builder-dfly7

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires anthropic, mcp, and includes scripts (resource) components.

What problem does it solve?

MCP Builder provides a structured, production-oriented guide to designing and implementing MCP (Model Context Protocol) servers so that LLMs can reliably interact with external APIs and services through well-designed tools, clear schemas, and robust transports.

Core Features & Use Cases

  • Design Guidance: Tool naming, discoverability, input/output schema patterns, annotations (readOnlyHint, destructiveHint, etc.), and actionable error messages to make tools agent-friendly.
  • Language Guides: Opinionated, runnable examples and patterns for Python (FastMCP + Pydantic) and TypeScript (MCP SDK + Zod) including project layouts, validation, and testing.
  • Evaluation & Harness: An evaluation guide and scripts to create stable, read-only multi-hop questions and run an automated evaluation harness against your MCP server using stdio or HTTP transports.
  • Use Case: Wrap a third-party API (e.g., GitHub or a CRM) with a typed MCP server exposing search and resource tools, then run the provided evaluation harness to validate LLM usability.

Quick Start

Generate a FastMCP or TypeScript MCP server that exposes search and read-only tools for your API and run the included evaluation script to validate tool usability.

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 for LLMs to interact with external APIs?

To build an MCP server, use Python with FastMCP and Pydantic or TypeScript with the MCP SDK and Zod. This Skill provides structured guidance for tool design, schema validation, and transport selection to expose external APIs to LLMs.

What is the best way to validate MCP tool usability before deploying?

The best way to validate MCP tool usability is to run an automated evaluation harness. This Skill includes scripts and examples to test your MCP server using stable, read-only multi-hop questions via stdio or streamable HTTP transports.

Does FastMCP work with TypeScript or do I need the MCP SDK?

FastMCP is for Python implementations, while TypeScript requires the MCP SDK. This Skill provides opinionated, runnable examples and patterns for both stacks, including project layouts, validation, and testing configurations.

How do I structure input schemas for MCP tools using Pydantic or Zod?

Structure input schemas using Pydantic in Python or Zod in TypeScript. This Skill guides tool naming, discoverability, and schema patterns, including annotations like readOnlyHint and destructiveHint to make tools agent-friendly.

Can I use stdio transport for an MCP server or should I choose streamable HTTP?

You can use either stdio or streamable HTTP transports for an MCP server. This Skill offers transport selection guidance and includes an evaluation harness capable of testing your tools across both transport methods.

Why does my LLM struggle to use my external API wrapper without clear error messages?

Your LLM may struggle without actionable error messages and clear schemas. This Skill helps you design agent-friendly MCP tools with proper naming, input/output validation, and robust error handling to ensure reliable LLM interaction.