mcp-docs-server
Search and fetch MCP protocol documentation using BM25 search with weighted scoring and stemming.
README
MCP Docs Server
A Model Context Protocol (MCP) server for searching MCP protocol documentation. Built with TypeScript using the official MCP SDK.
Features
- BM25 Search - High-quality ranked search using BM25 algorithm with Porter stemming
- Fast Startup - Indexes document titles at startup, fetches content on-demand
- Markdown Aware - Weights headers, code blocks, and links for better relevance
- stdio Transport - Runs as a local process for easy integration
Installation
Quick Start (Recommended)
Add to your MCP client configuration (Claude Desktop, Kiro, etc.):
{
"mcpServers": {
"mcp-docs-server": {
"command": "npx",
"args": ["-y", "@praveenc/mcp-docs-server"]
}
}
}
That's it! The server will be downloaded and run automatically.
Global Install
npm install -g @praveenc/mcp-docs-server
Then configure your MCP client:
{
"mcpServers": {
"mcp-docs-server": {
"command": "mcp-docs-server"
}
}
}
Tools
search_mcp_docs
Search MCP protocol documentation with ranked results.
Parameters:
query(string, required): Search query stringk(number, optional): Maximum results to return (default: 5). Returns fewer if fewer documents match.
Example:
{
"query": "transports",
"k": 5
}
Returns:
[
{
"url": "https://modelcontextprotocol.io/specification/2025-11-25/basic/transports.md",
"title": "Transports",
"score": 15.725,
"snippet": "MCP uses JSON-RPC to encode messages..."
}
]
fetch_mcp_doc
Fetch full document content by URL.
Parameters:
uri(string, required): Document URL from modelcontextprotocol.io
Example:
{
"uri": "https://modelcontextprotocol.io/specification/2025-11-25/server/tools.md"
}
Returns:
{
"url": "https://modelcontextprotocol.io/specification/2025-11-25/server/tools.md",
"title": "Tools",
"content": "# Tools\n\nTools enable servers to expose executable functionality..."
}
Testing with MCP Inspector
npx @modelcontextprotocol/inspector npx -y @praveenc/mcp-docs-server
Development
Clone the repository for local development:
git clone https://github.com/praveenc/mcp-docs-server.git
cd mcp-docs-server
npm install
Commands
# Run in development mode
npm run dev
# Run tests
npm test
# Type check
npm run typecheck
# Build
npm run build
# Test with MCP Inspector (development)
npm run inspect:dev
Local MCP Client Config (Development)
{
"mcpServers": {
"mcp-docs-server": {
"command": "npx",
"args": ["tsx", "/path/to/mcp-docs-server/src/index.ts"]
}
}
}
Architecture
src/
├── index.ts # MCP server entry point
├── config.ts # Configuration
├── tools/
│ └── docs.ts # search_mcp_docs, fetch_mcp_doc
└── utils/
├── cache.ts # Document caching
├── doc-fetcher.ts # HTTP fetching & HTML parsing
├── indexer.ts # BM25 search index
├── logger.ts # Logging utilities
├── stopwords.ts # Stop words list
├── text-processor.ts # Text utilities
└── url-validator.ts # URL validation
Search Algorithm
The search uses BM25 (Best Matching 25) with enhancements:
- Porter Stemming - Matches word variants (e.g., "running" → "run")
- Bigrams - Captures phrase matches (e.g., "tool input")
- Weighted Scoring:
- Title matches: 3-8x boost
- Header matches: 4x weight
- Code blocks: 2x weight
- Link text: 2x weight
- Domain Term Preservation - Technical terms like "mcp", "json", "stdio" are not stemmed
License
MIT
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。