Agentic Research

Fetch

90,704 stars

Official MCP reference server: fetches a given URL and converts web content to LLM-friendly Markdown; also ships a fetch prompt template.

First published:2024-11-22
data as of:
2026-09-30
Licence:
MIT / Apache-2.0(repo 正在授權轉移,見 repo LICENSE)
Transport:
stdio
Primitives exposed:
tools / prompts

What it can touch

Makes outbound HTTP(S) requests to any URL, respecting robots.txt by default (user-agent and proxy are customizable). It does not write local files, but whatever it fetches lands directly in the model context.

Risk

Risk runs both ways: fetched pages are untrusted input, so a malicious page can prompt-inject the model; and it is an exfiltration channel — an injected agent can encode sensitive data into URL parameters and send it to an attacker's server.

Install prompt

從官方參考 server 倉庫的 Fetch 目錄取得說明,把 README 的客戶端設定片段(以 uvx 啟動 mcp-server-fetch)加進你的 MCP 客戶端設定檔,然後重啟客戶端。

Paste this to your agent. It deliberately contains no tool subcommands — check the tool's official docs for commands.

Official source Source(90,704 ★)
Compatible with