Skip to main content

How we rank.

We’re a directory, not an ad network. Here’s exactly what goes into a ranking — and what doesn’t.

Where we are today: tools are ranked by our editor-assigned popularity score blended with third-party ratings; skills are ranked by live GitHub stars, with a separate “Rising” track for 7-day star growth. The full weighted system below is the roadmap we’re building toward as the community grows. No ranking position has ever been for sale.

In one paragraph

We track AI tools and agent skills. Tools rank by a blend of community upvotes, third-party ratings, and editorial review. Skills rank by GitHub stars, 7-day star growth, and trust level (verified maintainer / signed releases). We never accept payment to change a ranking. We do accept sponsorships — they’re labeled Sponsoredand never appear in the “Top” lists.

Ranking AI tools

Every tool gets a popularity score from 0–100. The score is recalculated nightly from these inputs:

40%
Community upvotes
Real, signed-in upvotes from registered users. One per account.
25%
Third-party ratings
Aggregated from G2, Capterra, Product Hunt, Trustpilot. Cached and refreshed weekly.
20%
Editorial review
Our team researches each tool that hits the directory. Tools scoring 75+ get a “Popular” badge — computed from the score, never sold.
10%
Recency / activity
Last commit, last release note, last public update. Stagnant tools decay.
5%
Search-result CTR
How often visitors who search a category click through to the tool. Helps surface intent matches.

Ranking agent skills

Skills are open source, so we lean on signals you can verify yourself by visiting the repo:

35%
GitHub stars
Total stars on the canonical repo. Reflects long-term reputation.
30%
7-day star growth
New stars in the last week. This is what powers the “Rising skills” carousel — fresh momentum, not legacy popularity.
15%
Trust level
Verified maintainer (Anthropic / Vercel / etc.), signed releases, and presence in official starter packs all count.
15%
Install activity
When available — npm downloads, MCP registry installs, Claude Skills installs.
5%
Recency
Last commit / last release. Skills that haven’t shipped in 6+ months get visually de-emphasized.

What we deliberately don’t do

  • Sell ranking positions. Sponsorship buys you the Sponsored slot, period.
  • Pass off editorial reviews as user reviews. Our team seeds short reviews distilled from public user sentiment (G2, Trustpilot, Reddit) so new pages aren’t empty — every one is labeled Editorial, and none of them move a ranking. Details on the disclosure page.
  • Use referral revenue as a ranking signal. We have affiliate links; they don’t move position.
  • Auto-rank by recency. New tools have to earn their spot.

About the save counts

The 👍 number on each tool combines real saves from signed-in users with a baseline computed from the tool’s popularity score. The baseline exists because raw counts on a young site would read 0–3 for everything; it’s deterministic, identical for every visitor, and real saves stack on top. Full explanation on the disclosure page.

Think a ranking is wrong?

We’d rather hear it. If you’re a maintainer, customer, or just a visitor who thinks something is mis-placed, write us. We read every note and adjust when the signal warrants it.

Dispute a ranking

Last updated July 24, 2026.