Fetch

What is Fetch?

Fetch is a Model Context Protocol server that enables LLMs to retrieve and process content from web pages by converting HTML to markdown for easier consumption. It provides a fetch tool and a fetch prompt for fetching URLs from the internet.

How to use Fetch?

Install Fetch via uvx (recommended), pip, or Docker. Configure it in your MCP client (e.g., Claude.app or VS Code) with the appropriate command and arguments. Use the fetch tool with a required url and optional max_length, start_index, and raw parameters to control content extraction. Paginate long pages by adjusting start_index.

Key features of Fetch

• Fetches URL content and converts to markdown
• Paginates results with start_index parameter
• Obey robots.txt by default for tool calls
• Customizable user-agent via --user-agent argument
• Supports proxy configuration with --proxy-url
• Raw content mode available with raw=true

Use cases of Fetch

• Retrieving and summarizing current web articles
• Reading long documentation in chunks to find specific info
• Feeding web data to AI models for analysis or answering questions
• Fetching reference material and converting to LLM-friendly format

FAQ from Fetch

What parameters does the fetch tool accept?

The fetch tool requires a url (string) and optionally accepts max_length (integer, default 5000), start_index (integer, default 0), and raw (boolean, default false).

Can the Fetch server access internal IP addresses?

Yes. The README warns that the server can access local/internal IP addresses and may represent a security risk. Exercise caution to avoid exposing sensitive data.

Does Fetch respect robots.txt?

By default, it obeys robots.txt for tool calls (model-initiated requests) but not for user-initiated prompts. You can disable this behavior with the --ignore-robots-txt argument.

What are the installation options for Fetch?

Fetch can be run via uvx (recommended), installed with pip and executed as python -m mcp_server_fetch, or used as a Docker container mcp/fetch.