mcp-builder

Guide MCP server design, implementation, and evaluation in TypeScript or Python.

Updated Mar 16, 2026
One-click install
npx skills add https://github.com/zlovexforever/zlx-claude-skills --skill mcp-builder-zlovexforever
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-builder
Source: https://github.com/zlovexforever/zlx-claude-skills/tree/main/mcp-builder
Command: npx skills add https://github.com/zlovexforever/zlx-claude-skills --skill mcp-builder-zlovexforever

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 Model Context Protocol (MCP) servers so LLM agents can reliably use external APIs through well-specified tools.

Core Features & Use Cases

  • Tool and workflow design guidance: Plan tool naming, coverage (API endpoints vs workflows), context management, and pagination strategies so agents find the right actions quickly.
  • Protocol- and SDK-informed implementation process: Follow a structured, phase-based roadmap for MCP server development in both TypeScript (TypeScript SDK) and Python (FastMCP/Python SDK).
  • Evaluation creation for real-world performance: Generate stable, read-only, complex evaluation questions (including an XML format) to test whether LLMs can accomplish tasks using only your MCP tools.
  • Best-practice references: Use bundled documentation pointers for MCP best practices and language-specific implementation patterns.

Quick Start

Implement an MCP server that integrates your target external service and then create a small evaluation set to verify tools are understandable, correctly typed, and effective for multi-step tasks.

Frequently Asked Questions about mcp-builder

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

FAQPage Schema
How do I build a Model Context Protocol server that LLM agents can use reliably?

To build a reliable Model Context Protocol server, follow a structured phase-based roadmap for TypeScript or Python, define clear tool schemas, and implement pagination strategies so agents can execute external-service tasks effectively.

How do I design MCP tool schemas with pagination for external APIs?

Design MCP tool schemas with pagination by following protocol-specific annotation conventions, providing robust descriptions, and using bundled framework documentation patterns to guide tool naming and context management for external APIs.

Can I implement an MCP server using FastMCP and the Python SDK?

Yes, you can implement an MCP server using FastMCP and the Python SDK, following the same structured development phases and best-practice references available for the TypeScript SDK to ensure protocol compliance.

What is the best way to test if LLM agents can accomplish tasks using my MCP tools?

The best way to test MCP tools is to generate stable, read-only evaluation cases that output an XML format with qa_pair entries, verifying whether LLMs can complete complex multi-step tasks using only your tools.

Why do my MCP tools fail to handle multi-step external-service tasks correctly?

MCP tools fail multi-step tasks when they lack robust descriptions, proper pagination, or stable evaluation cases, making it difficult for LLM agents to find the right actions and execute external-service workflows reliably.

Does the Model Context Protocol specification require specific error message guidance for tool design?

Yes, MCP specification review requires following error-message guidance and tool/annotation conventions during tool design, ensuring that LLM agents receive clear feedback when executing external-service tasks.