MCP Server Development Guide
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md and install @user_76b490f9/aaa-new.
About this skill
Problem
Building an MCP server is not just about calling external APIs. The harder question is whether an LLM can reliably discover the right tool, pass valid parameters, handle pagination and errors, and consume structured output. This skill turns that process into a staged workflow, reducing the gap between raw API documentation and a usable agent tool.
Workflow and boundaries
- Research and planning: review the MCP specification, best practices, and SDK docs before deciding on API coverage, tool naming, and context management.
- Implementation: prefer
TypeScriptorPythonfor remote or local servers, define input schemas withZodorPydantic, and addoutputSchema,structuredContent, and tool annotations where supported. - Testing and evaluation: verify the server through builds, syntax checks, and
MCP Inspector, then create 10 independent, read-only, complex, and verifiable evaluation questions.
Best for engineers wrapping external services into MCP tools. If the target API is unstable, or clients differ significantly in code execution and streaming support, validate compatibility first.
Use Cases
- Wrap an internal Git API into MCP tools so agents can create issues and list repositories.
- Design remote server tool naming, input schemas, pagination, and actionable errors per MCP spec.
- Implement external service tools in TypeScript with Zod, then verify them in MCP Inspector.
- Generate 10 read-only, complex, verifiable evaluation questions before shipping the server.
Best For
- Backend engineers integrating enterprise APIs into LLM agents who need tools that models can call reliably.
- Platform engineers building remote MCP services focused on Streamable HTTP, stateless JSON, and pagination.
- TypeScript developers implementing agent tools who need Zod schemas and structuredContent support.
- QA engineers evaluating agent tool quality who need read-only, verifiable test questions.
Related Skills
An AI workflow that structures personal knowledge bases, IMA, Yuque, and Feishu content into approved personas for paid WeChat mini-agent deployment.
Local workflow memory with matching and SOP updates.
An OpenClaw live streaming executor that initializes TRTC streaming, starts a real-time dashboard, generates viewer URLs, and continuously reports live events.
Breaks down physical supply chains for super-trends to identify second- and third-layer bottlenecks, runs valuation and reverse checks, and maintains trackable reports.