Web Page to Markdown Tool

The Web Page to Markdown Tool fetches one or more web pages by URL, extracts the main article content (stripping navigation, ads, and other noise), and returns an array of results as clean Markdown.
It uses a headed Chromium browser under the hood, which means it works on JavaScript-rendered pages, follows redirect chains (including Google News links), and can dismiss common consent/cookie banners automatically.
This tool is ideal for workflows that need to read and summarise the actual content of a URL found in a prior search.

Inputs

• urls ¹ (string[]): One or more full URLs of web pages to fetch (e.g., ["https://www.example.com/article"])
• requireToolUse (boolean, user config): Require tool use (forces the LLM to always call this tool; default: true)

  Note: When enabled, the node will retry tool calls up to the number of times set in Max Tool Retries in the AI Data Processing Node.

---

(1) Provided by the AI Data Processing Node as a result of processing its input, typically one or more URLs returned by a preceding search tool.

Outputs

• title: The page title
• url: The final URL after any redirects
• content: The main article body converted to Markdown
• excerpt: A short description extracted from the page's og:description or meta[name="description"] tag

Example Output:

[
    {
        "title": "Aliens might exist, but there are three reasons why they're not visiting us",
        "url": "https://theconversation.com/aliens-might-exist-but-there-are-three-reasons-why-theyre-not-visiting-us-283984",
        "content": "## Aliens might exist, but there are three reasons why they're not visiting us\n\nThe universe is vast, and the possibility of life elsewhere remains open...",
        "excerpt": "If extraterrestrial life exists, there are strong scientific reasons to think it may never reach Earth."
    }
]

Node Type

• ai-tool
• toolSubtype: web-page-to-markdown

Notes

• Redirect support — the tool waits for the full redirect chain to settle before extracting content, so short links, Google News read/ links, and similar forwarded URLs are handled transparently.
• Consent wall handling — common cookie/consent dialogs (including Google's consent screen) are automatically dismissed before the page content is read.
• Bot-detection mitigation — the browser launches with --disable-blink-features=AutomationControlled and masks navigator.webdriver, making it appear as a regular browser to most bot-detection systems.
• Browser path — in the Windows desktop (Tauri) app, a browser executable path must be set in the application Settings before this tool can be used.