Skip to main content
Glama
mcp-docs-server

mcp-docs-server

by praveenc
README.md
# MCP Docs Server

A Model Context Protocol (MCP) server for searching MCP protocol documentation. Built with TypeScript using the official MCP SDK.

[![npm version](https://img.shields.io/npm/v/@praveenc/mcp-docs-server.svg)](https://www.npmjs.com/package/@praveenc/mcp-docs-server)
[![MCP Registry](https://img.shields.io/badge/MCP-Registry-blue)](https://registry.modelcontextprotocol.io/v0.1/servers/io.github.praveenc%2Fmcp-docs-server/versions/0.1.0)

## 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.):

```json
{
  "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

```bash
npm install -g @praveenc/mcp-docs-server
```

Then configure your MCP client:

```json
{
  "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 string
- `k` (number, optional): Maximum results to return (default: 5). Returns fewer if fewer documents match.

**Example:**

```json
{
  "query": "transports",
  "k": 5
}
```

**Returns:**

```json
[
  {
    "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:**

```json
{
  "uri": "https://modelcontextprotocol.io/specification/2025-11-25/server/tools.md"
}
```

**Returns:**

```json
{
  "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

```bash
npx @modelcontextprotocol/inspector npx -y @praveenc/mcp-docs-server
```

## Development

Clone the repository for local development:

```bash
git clone https://github.com/praveenc/mcp-docs-server.git
cd mcp-docs-server
npm install
```

### Commands

```bash
# 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)

```json
{
  "mcpServers": {
    "mcp-docs-server": {
      "command": "npx",
      "args": ["tsx", "/path/to/mcp-docs-server/src/index.ts"]
    }
  }
}
```

## Architecture

```text
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

TDQS

A3.7/5.0

Scored across 2 tools

Disambiguation5/5

The two tools have clearly distinct purposes: one fetches a specific document by URL, the other searches across documents. No overlap in functionality.

Naming Consistency5/5

Both tool names follow a consistent verb_noun pattern (fetch_mcp_doc, search_mcp_docs) and use snake_case uniformly.

Tool Count3/5

With only two tools, the server is minimal but still covers the core operations of searching and fetching documentation. However, it feels slightly thin for a documentation server.

Completeness3/5

The server allows searching and fetching full documents, but lacks operations like listing all available documents or browsing by category, which are common in documentation servers.

Maintenance

ActivityInactive
ResponsivenessNo recent community issues