Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

unique-mcp-builder-test

bởi uniquevme

Web & Frontend Development
573Lượt tải do clawskills.sh công bố
0Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external.

Mô tả

Kỹ năng này làm gì

A guide for building MCP (Model Context Protocol) servers that expose external APIs and services as tools for LLMs. Covers TypeScript and Python implementations across four phases: research, implementation, testing, and evaluation creation.

Follows a four-phase workflow from API research through LLM-facing evaluation, reducing the common outcome of MCP servers that compile but agents cannot use effectively.

Khi nào nên dùng

  • Wrapping a third-party REST API so an AI agent can call it as a tool
  • Building a local stdio MCP server for development and testing workflows
  • Designing paginated, structured tool responses for complex data APIs
  • Adding actionable error messages so agents can recover from failures
  • Writing evaluation question sets to verify an MCP server works correctly with LLMs

Quy trình ví dụ

ĐẦU VÀO

User asks: Build a TypeScript MCP server that wraps the GitHub REST API

AGENT
  1. 1

    Fetches MCP specification and TypeScript SDK documentation

  2. 2

    Reviews GitHub API endpoints and selects priority operations to cover

  3. 3

    Sets up TypeScript project with Zod input schemas and MCP SDK

  4. 4

    Implements tools with consistent naming, pagination, and actionable error messages

  5. 5

    Creates 10 read-only evaluation questions and verifies answers against live API

ĐẦU RA

A deployable TypeScript MCP server with GitHub tools, structured output schemas, and an XML evaluation file

Cài đặt

clawhub install uniquevme/unique-mcp-builder-test

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết