mcp-readability

MCP server that fetches webpages and extracts clean article content with Mozilla Readability.

It exposes one MCP tool, readability, which accepts a list of URLs and returns structured article results with title, excerpt, site name, byline, plain text, cleaned HTML content, and length. Failed URLs are returned as per-URL errors instead of failing the whole request.

Requirements

• Node.js 18+
• Network access from the MCP server process to the URLs you want to parse

Install

Run directly with npx:

npx @khlesk/mcp-readability

Or install globally:

npm install -g @khlesk/mcp-readability
mcp-readability

MCP Client Config

Most MCP clients launch stdio servers with a command and args. With npx:

{
  "mcpServers": {
    "readability": {
      "command": "npx",
      "args": ["@khlesk/mcp-readability"]
    }
  }
}

With a global install:

{
  "mcpServers": {
    "readability": {
      "command": "mcp-readability"
    }
  }
}

Tool

readability

Extracts readable article content from one or more webpages.

Input:

{
  "urls": ["https://example.com/article"]
}

Successful result item:

{
  "url": "https://example.com/article",
  "ok": true,
  "title": "Article title",
  "excerpt": "Short summary or excerpt",
  "siteName": "Example",
  "byline": "Author name",
  "textContent": "Plain article text...",
  "content": "<div>Cleaned article HTML...</div>",
  "length": 1234
}

Failed result item:

{
  "url": "https://example.com/article",
  "ok": false,
  "error": "Failure message"
}