Fetch MCP Server

Content

Fetch MCP Server is an official Model Context Protocol reference server for retrieving web content. It exposes a fetch tool that requests a URL, extracts page content, and returns markdown for easier model consumption. It can also return raw content and read long pages in chunks using a start index.

The upstream README documents uvx, pip, and Docker configuration. The PyPI package exposes the mcp-server-fetch command for MCP clients.

Source Review

• https://github.com/modelcontextprotocol/servers
• https://github.com/modelcontextprotocol/servers/tree/main/src/fetch
• https://github.com/modelcontextprotocol/servers/blob/main/src/fetch/README.md
• https://github.com/modelcontextprotocol/servers/blob/main/src/fetch/pyproject.toml
• https://pypi.org/pypi/mcp-server-fetch/json

These sources were reviewed on 2026-06-05. Prefer the live repository and PyPI metadata for current package version, command name, Python requirement, dependencies, tool behavior, robots.txt behavior, and setup guidance.

Features

• Fetch a URL and extract page content as markdown.
• Request raw content when markdown conversion is not desired.
• Limit returned content with a maximum character length.
• Continue reading long pages by passing a start index.
• Use optional proxy, user-agent, and robots.txt behavior settings.
• Run as an official reference MCP server through uvx, pip, or Docker.

Installation

For MCP clients that launch stdio servers:

{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
      "args": ["mcp-server-fetch"]
    }
  }
}

Restart the MCP client after adding the server.

Use Cases

• Ask Claude to retrieve a public documentation page and summarize it.
• Convert an HTML article or reference page into markdown for review.
• Let Claude inspect long pages in chunks when the first response is truncated.
• Use controlled web fetching in workflows where a general browser MCP server is unnecessary.
• Compare page content across public URLs during research.

Safety and Privacy

Web fetching can become a data exposure path if the server can reach internal services, private pages, or authenticated resources. Keep network access constrained, review proxy and user-agent settings, and consider URL allowlists or policy controls before enabling this server in sensitive environments.

Fetched URLs, response bodies, internal hostnames, prompts, tool arguments, and errors may become part of the model context. Avoid fetching confidential, customer, authenticated, or internal pages unless the data is approved for the current model session.

Duplicate Check

No dedicated modelcontextprotocol/servers Fetch entry, mcp-server-fetch package entry, or matching source URL was found in content/mcp. The existing Sequential Thinking entry is from the same official server monorepo but covers a separate server and package.