mcp-builder

Guides building MCP servers in TypeScript, Python, or C# with evaluation tooling.

1|Updated May 29, 2026
One-click install
npx skills add https://github.com/memasanz/agent-harness --skill mcp-builder-memasanz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-builder
Source: https://github.com/memasanz/agent-harness/tree/main/.github/skills/mcp-builder
Command: npx skills add https://github.com/memasanz/agent-harness --skill mcp-builder-memasanz

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Building Model Context Protocol servers that LLMs can actually use effectively requires careful tool design, naming, schema validation, and testing—this Skill provides a structured four-phase workflow covering research, implementation, review, and evaluation. ## Core Features & Use Cases - Four-Phase Development Workflow: Deep research and planning, implementation, review and testing, and evaluation creation for MCP servers. - Multi-Language Guidance: Reference guides for TypeScript (MCP SDK with Zod), Python (FastMCP with Pydantic), and C#/.NET (Microsoft MCP SDK with Azure patterns). - Evaluation Harness: Scripts to run XML-based QA evaluations against your MCP server via stdio, SSE, or HTTP transports using Claude. - Use Case: You need to expose your company's internal API to LLM agents. Use this Skill to design well-named tools with proper schemas, implement the server, and validate it with 10 complex evaluation questions. ## Quick Start Ask the agent to help you build an MCP server for your API, specifying your preferred language (TypeScript, Python, or C#) and the service you want to integrate.

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 in TypeScript?

Use the @modelcontextprotocol/sdk package and register tools with server.registerTool, providing a title, description, Zod inputSchema, and annotations. Build with npm run build and test using the MCP Inspector via npx @modelcontextprotocol/inspector.

What is the difference between stdio and streamable HTTP transports in MCP?

stdio runs the server as a local subprocess for single-user desktop scenarios with no network setup. Streamable HTTP is for remote, multi-client cloud deployments and requires authentication such as Entra ID or API keys.

Should I build a custom MCP server or use an existing Microsoft one?

Check Microsoft's existing servers first: Azure MCP covers 48+ Azure services, Foundry MCP handles agents and evals, and GitHub MCP covers repositories. Build a custom server only for internal APIs, third-party SaaS, or specialized domain logic.

How do I test whether my MCP server tools work well with LLMs?

Create an XML evaluation file with 10 complex, read-only questions having single verifiable answers, then run scripts/evaluation.py against your server. The harness reports accuracy, tool call counts, and agent feedback on tool quality.

What makes a good MCP tool description?

Tool descriptions must narrowly and unambiguously describe functionality, document parameters and return schemas, and include usage examples. Use snake_case names with a service prefix like github_create_issue, and set annotations such as readOnlyHint and destructiveHint.

Why should MCP evaluation questions avoid current-state data?

Answers must be stable and verifiable by direct string comparison, so questions based on dynamic state like open issue counts become incorrect over time. Use historical, closed data such as completed projects or merged pull requests from a fixed time window.