MCP-Web-Fetch

A Model Context Protocol (MCP) server for fetching webpages including html/pdf/plain text type content.

Available Tools

• fetch-web: Fetch URL and return content according to its content type.
  • url: URL to fetch
  • raw: Whether to return raw content

Installation

Using uv(recommended)

When using uv, no specific installation is needed. We will use uvx to directly run mcp-web-fetch.

Using pip

pip install mcp-web-fetch

After installation, you can run the server using

python -m mcp-web-fetch

Configuration

Add to your local client

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

or using docker(after publish done)

{
    "mcpServers": {
        "mcp-web-fetch": {
            "command": "docker",
            "args": [
                "run",
                "-i",
                "--rm",
                "ghcr.io/mcp-servers/fetch:latest"
            ]
        }
    }
}

or Fetch and run from remote repository

{
  "mcpServers": {
    "mcp-web-fetch": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/pathintegral-institute/mcp.science#subdirectory=servers/web-fetch",
        "mcp-web-fetch"
      ]
    }
  }
}

Customization

By default, the server will use a default user-agent for fetching web content. You can customize the user-agent by setting the user_agent in args

uvx mcp-web-fetch --user-agent "Your User-Agent"