mcp-api-reference

mcp-api-reference

An MCP server that automatically crawls, indexes, and serves API reference documentation for LLMs, enabling to search and retrieve endpoint details.

Category
访问服务器

README

mcp-api-reference

日本語版 (Japanese)

An MCP (Model Context Protocol) server that automatically crawls, indexes, and serves API reference documentation for LLMs.

Problem & Solution

When LLMs generate code that uses APIs, the following problems occur:

  • Hallucination: Generating non-existent endpoints or parameters
  • Stale information: Unable to reference the latest API specs not included in training data
  • Inefficient context consumption: Passing entire documentation wastes tokens

This tool automatically crawls, parses, and indexes API reference sites just by specifying a URL, then provides token-efficient search results via MCP tools.

MCP Tools

Tool Description Main Inputs
search_docs Full-text search by keyword query, api (optional), limit (1-20)
get_endpoint Get details of a specific endpoint api, endpoint (path), method
list_apis List available APIs and categories api (optional)

Requirements

  • Node.js >= 18.0.0

Installation

npm install
npm run build

Usage

Start as MCP Server

node dist/index.js

CLI Options

# Specify a custom config file
node dist/index.js --config ./my-sites.json

# Refresh cache for a specific API
node dist/index.js --refresh kintone

# Clear all cache
node dist/index.js --clear-cache
Option Short Description
--config <path> -c Path to custom site config file
--refresh <api-id> -r Re-fetch documentation for the specified API
--clear-cache - Clear all cache and exit

Claude Desktop Integration

Add the following to claude_desktop_config.json:

{
  "mcpServers": {
    "api-reference": {
      "command": "node",
      "args": ["/path/to/mcp-reference-doc/dist/index.js"]
    }
  }
}

Presets

kintone REST API

A parser for the kintone REST API is built in as a preset. No additional configuration required.

Backlog API

A parser for the Backlog API is also available as a preset. No additional configuration required.

SmartHR API

A parser for the SmartHR API is also available as a preset. No additional configuration required.

Custom Sites

You can add any API documentation site by specifying a JSON file with the --config option.

{
  "sites": [
    {
      "id": "my-api",
      "name": "My API",
      "baseUrl": "https://api.example.com",
      "crawl": {
        "startUrl": "https://api.example.com/docs",
        "includePatterns": ["https://api.example.com/docs/**"],
        "excludePatterns": [],
        "maxPages": 500,
        "delayMs": 500
      },
      "parser": {
        "type": "generic",
        "selectors": {
          "endpointContainer": ".endpoint",
          "method": ".http-method",
          "path": ".api-path",
          "title": "h3",
          "description": ".description",
          "parameters": ".parameters",
          "responseFields": ".response"
        }
      }
    }
  ]
}

Project Structure

src/
├── index.ts              # CLI entry point
├── server.ts             # MCP server initialization and tool registration
├── core/
│   ├── crawler.ts        # Web crawler with robots.txt support
│   ├── parser.ts         # Parser registry
│   ├── generic-parser.ts # Generic HTML parser (CSS selector-based)
│   ├── indexer.ts        # Full-text search index via MiniSearch
│   ├── store.ts          # Document store
│   ├── cache.ts          # TTL-based cache management
│   └── pipeline.ts       # Crawl → parse → index pipeline
├── tools/
│   ├── search-docs.ts    # search_docs tool
│   ├── get-endpoint.ts   # get_endpoint tool
│   └── list-apis.ts      # list_apis tool
├── presets/
│   ├── kintone/          # kintone preset
│   ├── backlog/          # Backlog preset
│   └── smarthr/          # SmartHR preset
├── formatters/
│   └── response.ts       # MCP response formatter
├── types/                # Type definitions
└── utils/                # Utilities (logger, glob, hash)

Development

# Build in watch mode
npm run dev

# Run tests
npm test

# Run tests (watch mode)
npm run test:watch

# Type check
npm run typecheck

# Lint
npm run lint

Tech Stack

  • @modelcontextprotocol/sdk - MCP server implementation
  • cheerio - HTML parser
  • minisearch - Full-text search engine (with Japanese tokenization support)
  • zod - Schema validation
  • tsup - TypeScript bundler
  • vitest - Test framework

Environment Variables

Variable Description Default
MCP_API_REF_CACHE_DIR Path to cache directory ~/.mcp-api-reference/cache/
MCP_API_REF_CONFIG Path to site config file -

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选