Prompt file imported from paulanunes85/super-powers (.github/prompts/create-mcp-server.prompt.md). Fill in {{server_name}}, {{language}}, {{transport}}, {{tools_description}} before use. Copyright stays with the author.

Create MCP Server

Scaffold a complete, production-ready Model Context Protocol (MCP) server project.

Input

Server name: {{server_name}} Language: {{language}} (TypeScript, Python, or C#) Transport: {{transport}} (stdio or HTTP — default: stdio) Tools to create: {{tools_description}} (describe each tool's purpose)

Instructions

1. Load the matching skill based on {{language}}:

   • TypeScript → read typescript-mcp-server-generator skill
   • Python → read python-mcp-server-generator skill
   • C# → read csharp-mcp-server-generator skill

2. Scaffold the project following the skill's structure:

   • Project config (package.json / pyproject.toml / .csproj)
   • Server entrypoint with transport setup
   • One file per tool with schema validation
   • README with setup and usage instructions
   • .gitignore appropriate for the language

3. Implement all requested tools from {{tools_description}}:

   • Each tool gets a descriptive name, clear description, and input schema
   • Use zod (TS), Pydantic (Python), or DataAnnotations (C#) for validation
   • Include error handling and structured responses
   • Return both content and structured output where supported

4. Add client configuration snippets in the README:

   • VS Code (.vscode/settings.json MCP entry)
   • Claude Desktop (claude_desktop_config.json)
   • Claude Code (claude mcp add command)

Rules

• Follow the loaded skill's patterns exactly — do not deviate
• Use the latest MCP SDK version for the chosen language
• Every tool must have a clear, descriptive name and description (3-4 sentences)
• Input schemas must validate all parameters — no untyped inputs
• Include a health check or list-tools test command
• Server name in kebab-case: my-mcp-server

Output

A complete project directory with all files ready to run. After scaffolding, show:

1. Project tree — all created files
2. Quick start — commands to install, build, and run
3. Test command — how to verify the server works
4. Registration — config snippet for at least one AI client