Documentation Assistant MCP Server

A Model Context Protocol (MCP) server implementation for scraping, indexing, and searching documentation using Server-Sent Events (SSE).

Features

• SSE-based communication
• Documentation scraping and indexing with background job support
• Full-text search capabilities with content segment highlighting
• Local caching of documentation and sitemaps
• Intelligent sitemap generation
• Built with TypeScript and Express
• Follows MCP specifications

Installation

npm install

Building

npm run build

Running the Server

# Start with default port (9031)
npm start

# Start with custom port
node dist/index.js --port=9006

# Start with custom cache directory
node dist/index.js --cache-dir=/path/to/cache

The server will start on port 9031 by default.

API Endpoints

• SSE Connection: GET http://localhost:9031/sse
• Message Endpoint: POST http://localhost:9031/messages

Available Tools

1. get_cache_info

Get information about the cache location and contents.

{
  // No parameters required
}

2. get_sitemap

Creates or retrieves a sitemap by crawling the site for metadata and link structure. Returns a job ID for tracking progress.

{
  url: string; // The URL of the documentation site
}

3. get_job_status

Returns the current status, progress, and results of a background job.

{
  jobId: string; // The ID of the job to check
}

4. list_jobs

Returns a list of all background jobs with their current status and progress.

{
  // No parameters required
}

5. scrape_docs

Scrapes and indexes pages matching a search query. Uses cached data if available unless forceScrape=true. Returns a job ID for tracking progress.

{
  url: string;      // The URL of the documentation to scrape
  query: string;    // The search query to match content
  maxResults?: number;  // Maximum number of matching results (default: 5)
  forceScrape?: boolean;  // Whether to force a new scrape or use cache (default: false)
}

6. search_docs

Search through previously scraped documentation across single or multiple URLs. Returns relevant matches with title, description, URL and content segments, sorted by relevance.

{
  url: string | string[];  // Single URL or array of URLs to search
  query: string;           // The search query
  maxResults?: number;     // Maximum number of results (default: 10)
  maxSegments?: number;    // Maximum number of content segments per result (default: 3)
}

Cache Directory Structure

The server uses a local cache directory to store scraped documentation and sitemaps. By default, it's located at:

.cache/
  |- {base64_url}_sitemap.json  // Sitemap cache
  |- {base64_url}.json         // Documentation content cache

You can configure the cache directory location using:

• Command line: --cache-dir=/path/to/cache
• Environment variable: DOCS_CACHE_DIR=/path/to/cache

Background Jobs

The server supports long-running operations through a background job system. Jobs can be:

• PENDING: Job created but not started
• RUNNING: Job is currently executing
• COMPLETED: Job finished successfully
• FAILED: Job encountered an error

Use the get_job_status tool to monitor job progress and retrieve results.

License

MIT