Copying a page into an AI brings a few thousand tokens of menu with it
Free, copy or save, and it keeps tables and code blocks intact.
This exists because of one specific annoyance. Select a page, copy it, paste it into Claude or ChatGPT, and you have just spent a chunk of your context window on navigation links, a cookie notice, a newsletter box and a footer sitemap.
So this pulls out the content and leaves the furniture behind. Tables stay tables, code blocks stay code blocks, and nested lists survive, which is where a lot of converters fall over and quietly flatten everything.
How to use it
- Open the page.
- Click the icon.
- Copy to clipboard, or save as a .md file.
- For several pages, give it a sitemap URL and it will convert them into one file.
What gets thrown away
Navigation, footers, cookie banners, newsletter prompts, share buttons, related-article rails and comment widgets. Anything that is the site rather than the article.
That is a judgement call, and occasionally it will be wrong. On a page that is genuinely a list of links, a directory or a category index, boilerplate removal has nothing to keep and the output will look thin. That is the tool working correctly on the wrong sort of page.
What it can access, and why
| Permission | Why it needs it |
|---|---|
| Reads the current tab | Only when you click the icon. |
| Runs a script on the page | To read the rendered content, since a lot of pages assemble their article with JavaScript. |
| Downloads | To save the .md file. |
| Other pages on the site (optional) | Only if you convert more than one page, which has to fetch them and your sitemap, so it is requested at that point rather than at install. |
Full detail in the privacy policy.
FAQ
How much does this actually save on tokens?
It depends entirely on the page, and the honest answer is that I will not give you a percentage I made up. On a content-heavy article the saving is modest. On a documentation page inside a large site, where the sidebar tree is longer than the article, it is substantial.
Do images come through?
As Markdown image links pointing at the original URLs, not as embedded files. For feeding an AI that is usually what you want, and for archiving it is not.
Does it work on a page behind a login?
For the single page, yes, because it reads what is on screen in your browser. The multi-page sitemap conversion does not carry your session, so those fetches will get the public version.
What about pages that are mostly JavaScript?
Fine, because it reads the page after rendering rather than the source. A page that shows nothing until scripts run will still convert correctly.