zcaceres/fetch-mcp
Free灵活获取 JSON、文本和 HTML 数据的 MCP 服务器。
FreeFree tier
Inputs: urlOutputs: text
About zcaceres/fetch-mcp
Fetch MCP Server is an open-source Model Context Protocol server for retrieving web content. It provides multiple tools to fetch URLs and return raw HTML, converted Markdown, plain text, JSON, readable article content (via Mozilla Readability), and YouTube video transcripts. It supports custom headers, character limits, start indices, and proxy configuration. Can be used as an MCP server (via client config) or as a standalone CLI tool. Environment variables allow setting default response limits and maximum response body size.
Key Features
Fetch and return raw HTML, JSON, plain text, or Markdown from any URL
Extract main article content using Mozilla Readability (strips ads, nav, boilerplate)
Retrieve YouTube video transcripts with support for multiple languages
Configurable HTTP headers, response length, start index, and proxy support
Works as an MCP server (added to client config) or as a standalone CLI tool
Environment variable settings for default character limit and maximum response bytes
Pros & Cons
Pros
- Supports multiple output formats: HTML, JSON, text, Markdown, readable articles, and YouTube transcripts
- Readability extraction strips unnecessary elements for clean article content
- Offers flexible parameters: custom headers, length limits, start index, proxy
- Free and open source with easy installation via npx or npm
- Can be used both as an MCP server and as a CLI tool
Cons
- Limited to single URL per request (no batch or crawl capabilities)
- YouTube transcript extraction may require yt-dlp for best results
- No built-in rate limiting or caching mechanisms
- Does not execute JavaScript, so dynamic content may not be captured
Best For
Web scraping and content extraction for AI agent workflowsExtracting readable article content from news and blog postsFetching YouTube video captions for transcription or analysisRetrieving JSON data from APIs for processing by language modelsFetching plain text or Markdown versions of web pages for summarization
FAQ
How do I install Fetch MCP Server?
Add it as an MCP server to your client configuration using npx with the command 'mcp-fetch-server', or install it globally via npm with 'npm install -g mcp-fetch-server' and use the CLI command 'mcp-fetch'.
What output formats are supported?
The tool supports raw HTML, Markdown, plain text (without HTML tags/scripts/styles), JSON, readable article content (via Mozilla Readability returned as Markdown), and YouTube video transcripts.
Can I use a proxy when fetching?
Yes, you can provide a proxy URL using the 'proxy' parameter in the tool call or the '--proxy' flag in the CLI.
How do I get a YouTube transcript in a non-English language?
Use the 'fetch_youtube_transcript' tool with the 'lang' parameter set to the desired language code (default is 'en'). For example, '--lang es' for Spanish.