Разработка и отладка MCP-серверов

ai-tooling

Mcp developer is a Claude Code skill that helps design, debug, and extend MCP servers and clients in TypeScript or Python, connecting AI systems to external tools and data sources via the Model Context Protocol (JSON-RPC 2.0). Reach for it when you need to wire Claude up to an external API or database through MCP, when your server fails protocol compliance checks and you can't pinpoint why, when you want to scaffold a new MCP project quickly, or when a transport layer is misbehaving in production. The skill covers the full cycle: project initialisation (`npx @modelcontextprotocol/create-server` or `pip install mcp`), tool schema definitions with Zod or Pydantic, transport setup (stdio, HTTP, SSE), resource and tool handler registration, compliance verification with MCP Inspector, and adding authentication with rate-limiting. The focus is implementing both the server and client sides of MCP — not generic REST integration (use fastapi-expert or typescript-pro for that). Output is production-ready implementation files, schema definitions, and transport configuration. Works well for teams building AI tooling on top of the MCP ecosystem; not the right fit if you need a plain WebSocket or REST endpoint without JSON-RPC 2.0 protocol constraints. MIT licence, author Jeff Allan, version 1.1.0.