MCP Server
Fetch MCP Server
Fetch web pages for an agent, convert HTML into Markdown, and continue through long responses with explicit character offsets.
What is Fetch MCP Server?
Fetch is an official MCP reference server for controlled web retrieval. Its single tool accepts a URL, converts HTML to Markdown by default, and provides length and offset controls for reading large pages incrementally. Because it can reach internal addresses, deployment boundaries matter.
Key capabilities
HTML to Markdown conversion
Transforms ordinary web pages into a compact format that language models can process efficiently.
Chunked retrieval
Uses maximum length and start offset controls to continue through content without returning an unbounded response.
Use cases
Retrieve documentation
Load a public page as Markdown for research, summarization, or downstream tool use.
Read long pages incrementally
Continue from a character offset until the relevant portion of a large resource is found.
MCP capabilities
Tools
fetch
Retrieve a URL and convert its HTML content to Markdown or return raw content.
Prompts
fetch
Prompt template for retrieving a URL as Markdown.
Connection and auth
- Python with uvx or pip, or Docker
Client configuration
Claude Desktop
{
"command": "uvx",
"args": ["mcp-server-fetch"]
}