defuddle-mcp
An MCP server that extracts clean Markdown or HTML content from web pages by stripping away ads, navigation, and clutter. It offers tools to process URLs or raw HTML, returning structured metadata alongside the main article content.
Ask AI about defuddle-mcp
Powered by Claude · Grounded in docs
I know everything about defuddle-mcp. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
defuddle-mcp
MCP server that extracts the main content from web pages using defuddle. Returns cleaned Markdown or HTML with metadata (title, author, description, etc.), stripping away ads, navigation, sidebars, and other clutter.
No authentication required — defuddle runs locally as a CLI tool.
Tools
| Tool | Description |
|---|---|
defuddle_url | Fetch a URL and extract its main content |
defuddle_html | Extract main content from raw HTML |
Both tools return a structured result with: title, author, description, domain, content, word_count, and other metadata fields. Set markdown=false to get HTML instead of Markdown.
Setup
Requires Python >= 3.10 and Node.js >= 18.
uv sync
The defuddle npm package is installed automatically on first server start.
Run
uv run python -m src.main
The server starts on http://127.0.0.1:8080/mcp.
Test
uv run python -m src.client