Linked Zread Provider

Linked Zread Provider is an MCP server that provides AI assistants with advanced web content reading and comprehension capabilities, enabling deep extraction and analysis of web page content with intelligent parsing for articles, documentation, and structured content. This MCP bridge goes beyond simple URL fetching by providing semantic understanding of page structure, extracting key information, and delivering clean, well-formatted content optimized for AI consumption.

When to Use This MCP Server

Connect this server when...

• You need AI assistants to deeply read and comprehend web articles, documentation, and long-form content
• Your workflow requires extracting clean, well-structured text from complex web pages with advertisements and navigation
• You want semantic parsing that understands article structure, headings, quotes, code blocks, and metadata
• You are building research workflows that need to distill key information from lengthy web documents
• You need to extract author information, publication dates, and article metadata alongside content

Consider alternatives when...

• You only need basic HTML fetching without intelligent content extraction (use a simple fetch server)
• Your web reading needs require JavaScript rendering or interactive page elements
• You need to process non-web content like PDFs or documents (use a document conversion server)

Quick Start

Copy
# .mcp.json configuration
{
  "mcpServers": {
    "zread": {
      "command": "npx",
      "args": ["-y", "@zread/mcp-server"],
      "env": {
        "ZREAD_API_KEY": "your-api-key"
      }
    }
  }
}

Connection setup:

1. Obtain an API key from the Zread developer portal
2. Ensure Node.js 18+ is installed on your system
3. Add the configuration above to your .mcp.json file with your API key
4. Restart your MCP client to activate the content reader

Example tool usage:

# Read an article
> Read and summarize the article at https://example.com/blog/ai-trends-2026

# Extract documentation
> Read the API documentation at https://docs.example.com/api/v2 and list all endpoints

# Analyze content
> Read this research paper page and extract the key findings and methodology

Core Concepts

Concept	Purpose	Details
Content Extraction	Clean text retrieval	Intelligent removal of navigation, ads, and boilerplate to extract the main article content
Semantic Parsing	Structure understanding	Recognition of headings, paragraphs, lists, code blocks, quotes, and their hierarchical relationships
Metadata Extraction	Article context	Pulling author names, publication dates, categories, and other article metadata from page markup
Readability Scoring	Content quality assessment	Measuring content clarity, reading level, and information density for quality filtering
Format Optimization	LLM-friendly output	Converting extracted content into Markdown format optimized for language model processing

Architecture:

+------------------+       +------------------+       +------------------+
|  Web Content     |       |  Zread MCP       |       |  AI Assistant    |
|  Articles/Docs   |<----->|  Server (npx)    |<----->|  (Claude, etc.)  |
|  (Internet)      | HTTPS |  + Content Parser| stdio |                  |
+------------------+       +------------------+       +------------------+
        |
        v
+------------------------------------------------------+
|  Fetch > Parse > Extract > Clean > Format > Return    |
+------------------------------------------------------+

Configuration

Parameter	Type	Default	Description
ZREAD_API_KEY	string	(required)	API key for authenticated access to the Zread content extraction service
output_format	string	markdown	Format for extracted content (markdown, plain, html)
extract_metadata	boolean	true	Whether to include article metadata (author, date, tags) in the response
max_content_length	integer	50000	Maximum character count for extracted content to limit very long articles
follow_pagination	boolean	false	Whether to follow pagination links to extract content spanning multiple pages

Best Practices

1. Use Zread for article-style content. The server excels at extracting clean content from blog posts, news articles, documentation pages, and similar structured content. For highly interactive or application-like web pages, a headless browser approach may be more appropriate.

2. Enable metadata extraction for research workflows. When building research pipelines, enable metadata extraction to capture author names, publication dates, and categories. This metadata helps the AI contextualize content and assess source credibility.

3. Set appropriate content length limits. Very long pages (academic papers, extensive documentation) can produce extracted content exceeding context window limits. Set max_content_length based on your model's capacity and the level of detail needed.

4. Combine with search for comprehensive research. Use a web search MCP server to find relevant URLs, then pass those URLs to Zread for deep content extraction. This two-step workflow produces better results than trying to extract insights from search snippets alone.

5. Cache frequently accessed content. If you repeatedly reference the same documentation or article, fetch it once and store the extracted content rather than re-fetching each time. This reduces API usage and improves response times for frequently consulted sources.

Common Issues

Extracted content is missing or incomplete. Some websites use JavaScript-heavy rendering that prevents server-side content extraction. Check whether the page renders its main content through JavaScript by viewing the page source. For JS-rendered content, use a web provider with headless browser capabilities instead.

Metadata extraction returns empty fields. Not all web pages include structured metadata. Pages without Open Graph tags, JSON-LD, or standard meta elements will have missing metadata. The server extracts what is available but cannot infer metadata that the page does not provide.

Content extraction includes unwanted elements. The semantic parser may occasionally include sidebar content, related article links, or footer text. Report persistent false positives to improve the extraction model. For critical workflows, review extracted content for accuracy before using it in downstream processes.