What is agent-centric design?

It is a design philosophy that focuses on building tools optimized for how AI models think—prioritizing workflows, managing context windows, and providing human-readable identifiers.

What is an MCP server?

An MCP (Model Context Protocol) server is a specialized service that provides tools and resources allowing LLMs to interact with external systems, APIs, and local data.

Which programming languages are supported by this skill?

This skill provides detailed guides and implementation patterns for both Python (using FastMCP) and Node/TypeScript (using the official MCP SDK).

How do I verify if my MCP server is working correctly?

The skill includes an Evaluation Phase (Phase 4) which guides you in creating 10 complex, realistic QA pairs to test the server's effectiveness in real-world scenarios.

Does this skill help with input validation?

Yes, it provides standardized patterns for using Pydantic in Python and Zod in TypeScript to ensure all tool inputs are strictly validated and documented.

MCP Builder

Name: MCP Builder
Author: Krosebrook

byKrosebrook

•

API Development

Architects high-quality Model Context Protocol (MCP) servers to enable seamless LLM interaction with external services and APIs.

The MCP Builder skill provides a comprehensive framework for developing robust Model Context Protocol (MCP) servers that allow AI agents to interact effectively with external data and tools. It guides developers through an agent-centric design process, emphasizing workflow-based tool creation, context optimization, and actionable error handling. Supporting both Python (FastMCP) and Node/TypeScript (MCP SDK), the skill ensures implementation follows industry best practices while providing a rigorous four-phase workflow covering research, systematic implementation, code quality review, and the creation of verifiable evaluation suites.

Key Features

01Comprehensive 4-phase development and evaluation framework

02Automated quality checklists for production-grade integrations

03Multi-language support for Python and Node/TypeScript SDKs

04Agent-centric tool design for high-impact AI workflows

051 GitHub stars

06Standardized input validation patterns using Pydantic and Zod

Use Cases

01Building custom tools to integrate proprietary internal APIs with Claude

02Consolidating fragmented API endpoints into cohesive agent-driven workflows

03Creating verifiable evaluation datasets to benchmark AI tool performance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add krosebrook/source-of-truth-monorepo mcp-builder

For use in Claude.ai and ChatGPT

Download Skill