Fetch MCP Server

A Model Context Protocol (MCP) server that fetches URLs and converts HTML/PDF to Markdown for LLMs. TypeScript/Node port of the original mcp-server-fetch2, communicating over stdio.

[!CAUTION] This server can access local/internal IP addresses and may represent a security risk. Exercise caution to ensure it does not expose sensitive data.

The fetch tool truncates the response; use the start_index argument to page through content in chunks until the model finds what it needs.

Available Tools

• fetch — Fetches a URL and extracts its contents as markdown (HTML) or text (PDF).
  • url (string, required): URL to fetch
  • max_length (integer, optional): Maximum characters to return (default: 5000)
  • start_index (integer, optional): Start content from this character index (default: 0)
  • raw (boolean, optional): Return raw content without markdown conversion (default: false)

Prompts

• fetch — Fetch a URL and extract its contents as markdown.
  • Arguments:
    • url (string, required): URL to fetch

Installation

Requires Node.js 18+.

git clone https://github.com/fxcl/fetch.git
cd fetch
npm install
npm run build

The server runs as node dist/index.js. package.json declares a bin field, so once published npx @fnnm/fetch will also work.

Configuration

Configure for Claude.app / Claude Desktop

Add to your Claude settings:

{
  "mcpServers": {
    "fetch": {
      "command": "node",
      "args": ["/abs/path/to/fetch/dist/index.js"]
    }
  }
}

{
  "mcpServers": {
    "fetch": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "mcp/fetch"]
    }
  }
}

Configure for VS Code

Add the following to .vscode/mcp.json in your workspace, or to your User Settings (JSON) via Ctrl+Shift+P → Preferences: Open User Settings (JSON). The mcp key is required when using mcp.json.

{
  "mcp": {
    "servers": {
      "fetch": {
        "command": "node",
        "args": ["/abs/path/to/fetch/dist/index.js"]
      }
    }
  }
}

{
  "mcp": {
    "servers": {
      "fetch": {
        "command": "docker",
        "args": ["run", "-i", "--rm", "mcp/fetch"]
      }
    }
  }
}

Behavior — robots.txt

The server obeys a site's robots.txt when the request comes from the model (via the fetch tool), but bypasses it when the request is user-initiated (via the fetch prompt). This behavior is fixed in the implementation; there is no CLI flag to toggle it.

Behavior — User-agent

The server sends one of two fixed user-agents depending on whether the request is model-initiated (tool) or user-initiated (prompt):

ModelContextProtocol/1.0 (Autonomous; +https://github.com/modelcontextprotocol/servers)

ModelContextProtocol/1.0 (User-Specified; +https://github.com/modelcontextprotocol/servers)

The user-agent is not configurable.

Behavior — Proxy

The server honors the HTTPS_PROXY / HTTP_PROXY environment variables (lowercase variants too), wired through an undici ProxyAgent. To set a proxy in mcpServers, add an env block:

{
  "mcpServers": {
    "fetch": {
      "command": "node",
      "args": ["/abs/path/to/fetch/dist/index.js"],
      "env": {
        "HTTPS_PROXY": "http://user:[email protected]:8080"
      }
    }
  }
}

Debugging

Use the MCP inspector:

npx @modelcontextprotocol/inspector node dist/index.js

Contributing

Pull requests welcome. For examples of other MCP servers and implementation patterns, see https://github.com/modelcontextprotocol/servers.

License

MIT. See LICENSE.