mcp-server

Design MCP servers for Claude with FastMCP or the TypeScript SDK.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/christophevg/c3 --skill mcp-server-christophevg
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-server
Source: https://github.com/christophevg/c3/tree/main/skills/mcp-server
Command: npx skills add https://github.com/christophevg/c3 --skill mcp-server-christophevg

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill eliminates the uncertainty and trial-and-error involved in designing an MCP server so Claude can safely use external tools, read resources, and apply prompts.

Core Features & Use Cases

  • Server architecture guidance: Choose between Tools, Resources, and Prompts based on what Claude needs to do and what data it must access.
  • Deployment decisioning: Select the best transport model (Remote HTTP, MCPB bundled local server, or local stdio prototypes) depending on distribution needs, access requirements, and authentication constraints.
  • Design patterns & implementation paths: Apply one-tool-per-action or search + execute patterns, then scaffold the server using FastMCP (Python) or the TypeScript SDK.
  • Testing, publishing, and safety: Validate with MCP Inspector, publish as a plugin/deployment, and follow a security checklist covering logging, input validation, TLS, and credential handling.

Quick Start

Ask the AI to design an MCP server that wraps your REST API and expose it over secure HTTP for use with Claude.

Frequently Asked Questions about mcp-server

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

FAQPage Schema
How do I build an MCP server that lets Claude invoke external tools and read resources?

You design an MCP server by selecting tool, resource, and prompt patterns, then scaffold it using FastMCP for Python or the TypeScript SDK. This approach eliminates bespoke integrations by letting Claude safely invoke external tools and read resources.

What is the best way to wrap a REST or GraphQL API for use with Claude?

The best way to wrap a REST or GraphQL API for Claude is to apply the one-tool-per-action or search and execute design patterns within an MCP server. You then expose these tools over your chosen transport model for Claude to invoke directly.

How do I choose between stdio, Remote HTTP, and MCPB bundled transports for MCP server deployment?

Choose your MCP server transport based on distribution needs and authentication constraints: use local stdio for prototypes, MCPB for bundled local servers, and Remote HTTP for secure distributed access requiring TLS and credential handling.

Can I use FastMCP and the TypeScript SDK to scaffold an MCP server with a large action catalog?

Yes, you can scaffold an MCP server handling a large action catalog using either FastMCP in Python or the TypeScript SDK. You apply design patterns like search and execute to manage the catalog efficiently before validating with MCP Inspector.

What security practices should I follow when deploying an MCP server over Remote HTTP?

When deploying an MCP server over Remote HTTP, you must enforce security practices including input validation, safe logging, TLS encryption, and proper secret handling. Following a security checklist ensures Claude interacts with external tools safely.

How do I validate the correctness of an MCP server before publishing it as a plugin?

You validate the correctness of an MCP server using MCP Inspector before publishing it as a plugin or deployment. This testing phase ensures your tool, resource, and prompt implementations function correctly within the Claude environment.