I built 9 free web tools and printed the entire backend on every page. What should number 10 be?

Over the last couple of weeks I've built nine small web tools. No signup, no API key, no "start your free trial". You paste a URL, you get a result.

There's a catch, and it's the interesting part: every tool prints the complete source code of its own backend, right there on the page. Not a snippet. Not pseudocode. The actual file that ran when you clicked the button.

I did that because I wanted the tools to be an argument, not a brochure. If I claim a backend is small, you shouldn't have to take my word for it.

Now I've run out of good ideas for number ten, which is why I'm writing this. Skip to the bottom if you just want to tell me what to build.

The nine

• URL to Markdown — paste a link, get clean Markdown for an LLM prompt, plus the token count before and after so you can see what the HTML was costing you.
• Page to JSON — one page as structured JSON: title, H1, description, text, every link with its anchor text, every image with its alt. Missing anchor and alt text get flagged rather than silently returned as empty strings.
• Website to JSON — same idea, but walks the sitemap instead of one page.
• URL to YAML — a page's metadata as YAML.
• llms.txt generator — builds an llms.txt from your sitemap.
• AI Crawler Check — which AI crawlers your robots.txt actually allows or blocks, citing the exact rule that decided each one. Also checks whether your content survives without JavaScript, because most AI crawlers don't render it.
• Link Preview Checker — what your Open Graph and Twitter cards really look like to a scraper.
• Website Image Gallery — every image across your site with its alt text, so you can find the ones missing it.
• Dead Link Checker — crawls your sitemap and reports links that don't answer 200.

All nine are here: hyperlambda.dev/tools

What "the whole backend" actually looks like

Here's URL to Markdown, in full. This is not an excerpt:

.arguments
   url:string
validators.mandatory:x:@.arguments/*/url
validators.url:x:@.arguments/*/url
http.get:x:@.arguments/*/url
html2markdown:x:-/*/content
   url:x:@.arguments/*/url
openai.tokenize:x:@http.get/*/content
openai.tokenize:x:@html2markdown
yield
   markdown:x:@html2markdown
   html_tokens:x:@openai.tokenize/@openai.tokenize
   markdown_tokens:x:@openai.tokenize

Fetch it, convert it, count the tokens twice, return three fields. The language is Hyperlambda, which is my own thing, so take the language advocacy with as much salt as you like — the point I care about is that a working tool can be this small, and that you can check.

Not all nine are that short. The dead link checker is about seventy lines, because deduplicating links and surviving hosts that refuse connections is genuinely more work. That page says so and prints all seventy.

They're also MCP tools

Every HTTP endpoint on this platform is automatically exposed as an MCP tool, filtered by role. So these aren't just web pages — an agent can call the same endpoints directly, and "build a tool" and "build an agent tool" stopped being two separate tasks.

That turned out to be the thing I use most. Half of these exist because I wanted Claude to be able to read a page properly.

The honest limits

Things I'd rather you hear from me than discover:

• No JavaScript rendering. Everything fetches server-rendered HTML. A client-side-rendered SPA comes back nearly empty — which is also exactly what most AI crawlers see, so arguably that's the useful answer.
• The dead link checker crawls 10 sitemap pages max, though it checks every link it finds on them.
• The image gallery caps out at 50 results.
• Some sites return 403/999 to anything that isn't a browser. That's a bot refusal, not a broken link, and the dead link checker separates those into their own bucket instead of calling them dead.
• These are free and shared. Please don't point a load test at them.

So — what should number 10 be?

This is the actual reason for the post. I'm better at building these than at thinking of them.

What I'm after is the thing you currently do by hand, or with a browser extension you don't trust, or by pasting into three different sites in sequence. Ideas I've had and rejected as too boring: another JSON formatter, another minifier, another base64 thing.

Some prompts, if it helps:

• What's the last web-related task where you thought "surely something does this for me" and nothing did?
• Is there a tool you use that's fine except for one missing feature?
• What would you want an AI agent to be able to do to a web page, that it currently can't?

Reply with whatever comes to mind. If I build yours I'll say so in the thread, and it'll be free and no-signup like the rest.

And if one of the nine is broken or wrong for your site, I'd genuinely rather know — tell me the URL you tried.