I Built an AI Tool Directory That Recommends Stacks Instead of Listing 45,000 Tools

allen — Wed, 12 Aug 2026 12:49:13 +0000

I got tired of AI tool directories. Not because they're bad, but because they all do the same thing.

There's An AI For That has 45,000 tools. Toolify has 27,000. You search "AI coding" and get 500 results. Which one do you actually pick? Nobody tells you.

The question indie developers actually ask is different: "What tools do I need to build my product?" Not "list me 500 coding tools" but "give me the 5 tools that work together to ship a SaaS by Sunday."

So I built BotAI.

What it does differently

Three things, specifically.

Tool stacks. Instead of categories, BotAI has curated stacks. "Build a SaaS in a weekend" gives you Bolt for app generation, Supabase for database and auth, Stripe for payments, Resend for email, Vercel for hosting. Five tools, $0-30/month, sorted by workflow step. Each step says why this tool and what the alternatives are.

Comparisons. Cursor vs Copilot. Vercel vs Netlify. Supabase vs PlanetScale. Side-by-side tables with pricing, features, and a one-line verdict. Not two separate listing pages but actual head-to-head comparisons.

Use-case guides. "Best AI tools for coding", "Best database for Next.js" — pages with FAQ sections that directly answer the questions people type into Google and ChatGPT.

The tech

Next.js 15, static generation, Tailwind CSS
Fuse.js for client-side search (no API cost at this scale)
Resend + Upstash Redis for the newsletter system
Creem handles the $99 Featured Listing payment
A Python pipeline scrapes Hacker News, RSS feeds, Product Hunt, and GitHub trending, then filters out articles (this was harder than expected — more on that below)

50 tools to start. All human-reviewed. I'd rather have 50 tools with real descriptions and honest pros/cons than 45,000 scraped entries.

SEO is changing — GEO is the new game

Here's what I spent the most time on: making sure the site gets cited by AI search engines, not just ranked by Google.

The reasoning is simple. Matt Wolfe, who ran FutureTools.io (one of the biggest AI directories), announced his traffic dropped 80%. His words: "People don't need a tools directory anymore. They can just ask ChatGPT."

He's half right. People don't need a list. But they still need answers. So I optimized for being the source AI engines quote:

Every tool stack page has a short, direct-answer paragraph at the top. When someone asks ChatGPT "what tools do I need to build a SaaS", that paragraph is designed to be quotable.
FAQ schema on every use-case page. Four questions, four answers, each self-contained.
llms.txt at the root listing all pages for AI crawlers.
Breadcrumbs, ItemList, SoftwareApplication JSON-LD across all page types.

The goal isn't to fight AI search. It's to be the thing AI search cites.

What I learned building this

The discovery pipeline was the hardest part. My initial scraper grabbed 20 items from Hacker News per run. 14 of them were news articles about AI tools, not the tools themselves. "Adobe's new image rotation tool" was a CreativeBloq article. "Reverse engineering a $1B Legal AI tool" was a security blog post.

I wrote a heuristic filter that checks URL patterns (blog/article paths get rejected), title patterns (sentence-style headlines get rejected), and domain (news sites and social platforms get rejected). It also strips "Show HN:" prefixes before checking for colons, because the colon rule was killing the highest-value signal: actual Show HN tool launches.

The result: 20 items down to 6 real tools. The other 14 were articles. Better to miss a tool than list a blog post.

What's next

Growing the catalog to 200+ tools via the pipeline (with human review). Shipping a weekly newsletter. Adding a Pro tier for featured listings.

If you're an indie developer, browse the stacks. If you've built an AI tool, submit it — free listings, reviewed by a person.