Show HN: Linkidex - save and sort the URLs you care about, in clean Markdown

I spend a lot of time turning web pages into source material for AI tools. Hacker News threads, product docs, private dashboards, old forum posts, help center articles, internal wikis, pricing pages, GitHub issues. The annoying part is rarely reading the page. It is getting the useful parts into ChatGPT, Claude, or Cursor without dragging along navigation, ads, login banners, cookie boxes, and random sidebar text.

The Hacker News post titled Show HN: Linkidex - save and sort the URLs you care about is a good example. Linkidex is a bookmark manager built around search, categories, and tags. The founder describes using it to avoid re-finding wikis, Jira epics, proposals, and other work links. It is a progressive web app, with a Rails back end and a React, TypeScript, and GraphQL front end. The post also mentions offline support, browser import and export, 2FA, and WebAuthn.

That is useful context, but if you are feeding the thread into an AI tool, raw copy and paste is messy. You want the title, the author note, the core feature list, maybe selected comments, and a format the model can parse cleanly.

I tested the thread with Web2MD, our Chrome extension for converting web pages to Markdown. The result was simple: open the page, click the extension, review the token count, copy the Markdown, or send it to an AI tool in one click.

Here is the kind of Markdown output you want from a Show HN page:

# Show HN: Linkidex - save and sort the URLs you care about

Source: Hacker News
URL: https://news.ycombinator.com/item?id=33153866

Linkidex is a bookmark manager for saving, searching, and organizing URLs.

The author says it was built after regular browser bookmarks and an existing Chrome extension became hard to manage at work. Linkidex searches across link titles, URLs, categories, and tags. Results open in a new tab.

Technical notes from the post:

- Progressive web app
- Mostly works offline
- Rails back end
- React, TypeScript, and GraphQL front end
- AWS deployment
- 2FA and WebAuthn support
- Browser bookmark import and export

That is not fancy. That is the point. Clean Markdown beats a pretty extraction when your next step is summarizing, comparing, drafting, or coding.

Why this thread is a useful test case

A Hacker News thread has a few extraction traps.

The page is mostly text, but it is not a plain article. There is the story title, metadata, a URL, the submitter text, nested comments, voting controls, timestamps, reply links, and HN's compact layout. If you copy the page manually, you usually get extra junk. If you use a reader mode, you may lose comment structure or useful metadata.

For the Linkidex thread, the interesting parts are split across the original post and the discussion. The post explains the product. The comments contain questions about bookmark workflows, privacy expectations, browser extension habits, and whether people still want standalone bookmark managers. If you are researching "Show HN: Linkidex - save and sort the URLs you care about" for product positioning, you probably want both the pitch and the objections.

Web2MD keeps enough structure to make the page usable in an AI prompt. I could paste the Markdown into Claude and ask for a concise product teardown. I could send it to Cursor and ask it to draft a comparison page for bookmark tools. I could also ask ChatGPT to extract recurring feature requests from the comments.

A useful second output looks more like research notes:

## Product positioning notes

Linkidex positions itself as a searchable bookmark manager for work links.

Pain points mentioned:

- Too many important URLs to remember
- Browser bookmarks do not work well across all browsers and devices
- Existing URL management extensions can become hard to use as lists grow
- Users want fast search across title, URL, category, and tags

Potential comparison points:

- Browser bookmarks
- Raindrop-style bookmark managers
- Read-it-later apps
- Internal company wiki search
- Personal knowledge bases

That is the form I want when I am moving from a live page to an AI workflow. Not a screenshot. Not a blob of HTML. Not a cleaned page that silently drops the bits I needed.

Where browser-side conversion matters

Jina Reader is strong for public pages. It is fast, simple, and easy to call from a URL. Firecrawl is stronger when you need crawling, scraping, and developer APIs. MarkDownload is a good browser extension if your main need is clipping a page into Markdown.

Web2MD overlaps with all three, but it wins in a different spot: pages your browser can see.

That distinction matters. Server-side tools cannot read your logged-in Notion page, private Linear issue, internal Confluence page, paid newsletter, SaaS dashboard, or anything behind a session unless you do extra authentication work. Sometimes that is not possible. Sometimes it is a bad idea.

Web2MD runs in Chrome, in your browser session. If you can view the page, Web2MD can usually convert it. The content does not need to be sent through a scraping server first. For private or paid content, that is the safer default.

The extension is also practical for small daily workflows. The free tier gives you 3 conversions per day without an API key. Pro is $9 per month if you need more. I like that tradeoff because many people do not need an industrial crawler. They need to convert a few pages a day and get clean input for an AI model.

The token counter is not a small feature

The built-in token counter is one of the reasons I keep using Web2MD instead of plain copy and paste.

When I tested the Linkidex HN thread, I did not just want Markdown. I wanted to know whether the whole thread fit into my model context. A long HN thread can balloon quickly, especially if comments are nested or people quote each other. The token counter lets you decide before you paste.

That changes the workflow. Instead of sending a huge page and hoping the model reads it, you can trim first. Maybe you keep the original post and top comments. Maybe you remove boilerplate. Maybe you split the thread into two prompts.

For AI tools, clean input is only half the job. Sized input is the other half.

Honest limits

Web2MD is Chrome-only right now. If you live in Firefox or Safari, that is a real limitation.

The free tier is also intentionally limited to 3 conversions per day. That is enough for testing and light use, but not enough for people doing heavy research, content operations, or daily AI-assisted analysis. Pro is $9 per month.

It is also not a crawler. If you need to crawl hundreds of pages, schedule jobs, or build a scraping pipeline, Firecrawl is probably the better fit. If you need a simple public-page reader URL, Jina Reader is excellent. If you want a lightweight Markdown clipper and already like MarkDownload, you may not need to switch.

But if your real workflow is "I am looking at this page in Chrome and I want clean Markdown for ChatGPT, Claude, or Cursor," Web2MD is built for that exact moment.

A better way to research Show HN posts

Show HN threads are useful because they contain both the maker's pitch and the market's immediate response. For Linkidex, the post explains a concrete bookmark problem: too many important URLs, spread across work tools and devices, with regular bookmarks failing to keep up.

That is exactly the sort of page I want to preserve as Markdown. It becomes searchable, quotable, and easy to pass into an AI tool. You can ask for a product summary, extract objections, compare it with competitors, or turn the discussion into research notes.

If you want to try the same workflow, install Web2MD, open the Linkidex Show HN thread, convert it to Markdown, and check the token count before sending it to your AI tool. The free tier is enough to test a few pages and see if the browser-side workflow fits how you work.