mcp-builder
Build MCP Servers for Claude Code
Auch verfügbar von: AutumnsGrove,92Bilal26,YYH211,anthropics,92Bilal26,davila7,ZhanlinCui,ArtemisAI,Azeem-2,Doyajin174,DYAI2025,Cam10001110101,ComposioHQ,mcp-use
This skill provides comprehensive guidance for building Model Context Protocol servers that enable Claude Code and Codex to interact with external APIs through well-designed tools. It covers TypeScript and Python implementations, best practices, and evaluation strategies.
Die Skill-ZIP herunterladen
In Claude hochladen
Gehe zu Einstellungen → Fähigkeiten → Skills → Skill hochladen
Einschalten und loslegen
Teste es
Verwendung von "mcp-builder". How do I set up a basic MCP server project in TypeScript?
Erwartetes Ergebnis:
To set up a basic MCP server project in TypeScript:
1. Initialize a new Node.js project: npm init -y
2. Install the MCP TypeScript SDK: npm install @modelcontextprotocol/server-sdk
3. Create your server file with tool definitions using Zod for input validation
4. Register tools with the server and define the transport (stdio or Streamable HTTP)
5. Export the server for use with your chosen transport
The skill provides complete examples for each step with code snippets.
Verwendung von "mcp-builder". What are the best practices for tool naming in MCP servers?
Erwartetes Ergebnis:
Best practices for MCP tool naming:
- Use consistent prefixes (e.g., github_create_issue, github_list_repos)
- Use action-oriented names that describe what the tool does
- Make names descriptive of the service being integrated
- Avoid version numbers in names
- Use clear, unambiguous names that help agents find the right tool quickly
Verwendung von "mcp-builder". How should I structure error messages in my MCP server?
Erwartetes Ergebnis:
Error messages should be actionable and guide agents toward solutions:
- Include specific suggestions for fixing the error
- Provide next steps the agent can take
- Use standard JSON-RPC error codes
- Make error messages descriptive enough for the LLM to understand what went wrong and how to recover
Sicherheitsaudit
SicherThis skill is a documentation and guide resource for building MCP servers. The static scanner detected patterns (network calls, environment variable access, shell commands) in reference documentation files containing code examples. These are all false positives - the skill contains no executable code, only markdown documentation with example code snippets demonstrating standard API integration patterns.
Risikofaktoren
⚡ Enthält Skripte
🌐 Netzwerkzugriff
📁 Dateisystemzugriff
🔑 Umgebungsvariablen
⚙️ Externe Befehle
Qualitätsbewertung
Was du bauen kannst
Build a new MCP server for a custom API
A developer wants to create an MCP server to integrate their company's internal API with Claude Code. They follow the step-by-step guide to plan, implement, and test their MCP server.
Learn MCP protocol best practices
A technical writer or product manager wants to understand how MCP servers work and what makes them effective for LLM integration.
Evaluate existing MCP server quality
An AI engineer wants to evaluate their MCP server implementation to ensure LLMs can effectively use the tools provided.
Probiere diese Prompts
Help me create a basic MCP server using the TypeScript SDK. I want to expose a simple API (like a todo list) with create, read, update, and delete tools. Show me the project structure and core implementation.
How do I add API key authentication to my MCP server? I need to read the API key from an environment variable and include it in requests to the external service.
Review my MCP server tool descriptions and suggest improvements. I want to make sure Claude Code can understand what each tool does and use them effectively. Here are my current tool definitions: [add tool definitions]
Help me create an evaluation framework for my MCP server. I want to test whether Claude Code can effectively use my tools to answer realistic questions about my API.
Bewährte Verfahren
- Prioritize comprehensive API coverage over high-level workflow tools to give agents flexibility
- Use explicit tool definitions with title, description, inputSchema, and annotations
- Design input validation with Zod (TypeScript) or Pydantic (Python) for runtime type safety
- Test your MCP server with evaluation frameworks to ensure LLMs can effectively use your tools
Vermeiden
- Avoid vague tool names that do not clearly describe the action performed
- Do not skip input validation - always validate user inputs with proper schemas
- Avoid returning overly verbose or unfiltered data from tools - design focused responses
- Do not use generic error messages - provide actionable guidance for recovery