mcp-builder

Guides building MCP servers with TypeScript or Python SDKs, tool design, and evaluations.

Updated Sep 10, 2026
One-click install
npx skills add https://github.com/joydai2026-del/skills --skill mcp-builder-joydai2026-del
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mcp-builder
Source: https://github.com/joydai2026-del/skills/tree/main/mcp-builder
Command: npx skills add https://github.com/joydai2026-del/skills --skill mcp-builder-joydai2026-del

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Building a Model Context Protocol server that LLMs can actually use well requires many design decisions—tool naming, schemas, pagination, transports, error messages, and evaluation—that are easy to get wrong without a structured process. ## Core Features & Use Cases - Four-Phase Workflow: Deep research and planning, implementation, review and testing, and evaluation creation for MCP servers. - Language-Specific Guides: Reference docs for TypeScript (Zod, registerTool, streamable HTTP/stdio) and Python (FastMCP, Pydantic) with complete working examples and quality checklists. - Evaluation Framework: Guidelines for creating 10 complex, read-only, verifiable QA pairs in XML format to test whether LLMs can use your server effectively. - Use Case: You need to connect Claude to the GitHub API. Follow the guide to design tools like github_create_issue with proper annotations, implement the server, then generate an evaluation file to verify agents can answer realistic questions with it. ## Quick Start Ask the agent to build an MCP server for your chosen API using the mcp-builder guide, starting with research and ending with an evaluation file.

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 an external API?▼

Follow a four-phase process: research the MCP spec and target API, implement tools with Zod (TypeScript) or Pydantic (Python) input schemas, test with MCP Inspector, then create evaluation questions. Use streamable HTTP transport for remote servers and stdio for local ones.

Should I use TypeScript or Python for an MCP server?▼

TypeScript is recommended for its high-quality SDK support, static typing, and strong linting tools. Python with FastMCP is also fully supported, offering decorator-based tool registration and Pydantic validation. Both have complete implementation guides with examples.

What naming conventions should MCP tools follow?▼

Use snake_case with a service prefix in the format {service}_{action}_{resource}, such as github_create_issue or slack_send_message. Servers are named {service}_mcp in Python and {service}-mcp-server in TypeScript, without version numbers.

How do I test whether my MCP server works well with LLMs?▼

Create 10 complex, read-only, independent evaluation questions with single verifiable answers in XML format. Run them with the evaluation script over stdio, SSE, or HTTP transport, then review accuracy, tool call counts, and agent feedback to iterate on tool design.

What makes a good MCP evaluation question?▼

Good questions are realistic, complex multi-hop tasks requiring multiple tool calls, with stable answers that will not change over time. Avoid keyword-searchable questions, dynamic current-state queries, and answers that are lists or ambiguous formats.

When should I use stdio versus streamable HTTP transport?▼

Use stdio for local integrations and command-line tools where the server runs as a subprocess of the client. Use streamable HTTP for remote servers, web services, and multi-client scenarios; SSE is deprecated in favor of streamable HTTP.