Fetch

Fetch MCP Server retrieves web content from URLs and converts HTML to markdown so you can read and process web pages with an LLM. It includes built‑in options to fetch content in chunks and extract just what you need, making web data more accessible for your workflows.

How to use

You use the Fetch MCP Server by running it as a local process and then sending fetch requests to retrieve and transform web content. The server exposes a fetch tool that takes a URL and returns the page content as markdown. You can specify how much content to return with max_length and where to start with start_index, so you can read pages in chunks until you locate the information you need.

How to install

Choose one of the supported installation methods below and run the corresponding commands to start the server.

# Install and run via uvx (recommended for quick setup)
# uvx will run the MCP server fetch tool directly
uvx

# Start the fetch server using uvx with the appropriate package
# This step assumes you have uvx available in your environment
uvx mcp-server-fetch

Install using Docker if you prefer containerized running. You pull the image and start the server in a container.

docker run -i --rm mcp/fetch

Install via Python pip if you want to run the server as a Python module.

pip install mcp-server-fetch
python -m mcp_server_fetch

Configuration and usage notes

Configure how your client connects to the server by selecting one of the available start methods. You can run the server from different environments and reference it from your MCP client configuration.

Security and customization

Be mindful that this server can access local/internal IP addresses. Exercise caution to avoid exposing sensitive data.

Robots.txt handling, user‑agent selection, and proxy usage can be customized via command line arguments when starting the server. You can disable robots.txt by adding the flag --ignore-robots-txt. You can customize the user agent by adding --user-agent=YourUserAgent. You can route traffic through a proxy with --proxy-url.

Debugging

If you are developing or troubleshooting, you can use an inspector tool to debug the MCP server. The inspector provides visibility into the server’s behavior and requests.

Examples of how to fetch a page

# Example call (conceptual, client-side usage depends on your MCP client)
# Fetch a URL and get content as markdown
# url: https://example.com
# max_length: 5000 (default)
# start_index: 0 (default)
# raw: false (default)

Available tools

fetch

Fetches a URL and extracts its contents as markdown, with options to limit length, start from a specific index, and return raw content if requested.