Fetch a web page's main content and convert it to clean Markdown — as an MCP server and a CLI.
日本語版: README.ja.md
gangue renders a URL with headless Chrome, strips navigation, ads, and other noise, extracts the main article, and returns clean Markdown. It exposes a single MCP tool, fetch_markdown, and also works as a one-shot CLI.
- One tool, one job:
fetch_markdown(url) -> Markdown. Nothing else. - Real rendering: uses headless Chrome, so JavaScript-rendered pages work.
- Clean output: readability-style main-content extraction, converted to Markdown. If extraction finds too little, gangue falls back to the whole page, so you still get the content.
- MCP server or CLI: run it as a stdio MCP server, or pass a URL to print Markdown to stdout.
- Self-healing browser: the browser is launched once and reused; if it dies, gangue restarts it and retries.
- Google Chrome or Chromium installed on the system (gangue does not bundle a browser).
- For the
npxinstall method: Node.js.
Register gangue with your MCP client. With Claude Code:
claude mcp add --transport stdio gangue -- npx -y gangue@latestOr in .mcp.json:
{
"mcpServers": {
"gangue": {
"command": "npx",
"args": ["-y", "gangue@latest"]
}
}
}@latest makes npx fetch the newest published version on each launch, so you stay up to date.
cargo install --git https://github.com/dominion525/gangueOnce registered, the fetch_markdown tool is available:
fetch_markdown(url: string) -> string— returns the page's main content as Markdown.
gangue https://example.com/Prints the Markdown to stdout. With no argument, gangue runs as an MCP stdio server.
URL
→ headless Chrome renders the page
→ readability extracts the main content (relative links resolved to absolute)
→ HTML is converted to Markdown, with the title as a leading heading
→ Markdown
Licensed under either of MIT or Apache-2.0 at your option.