The VibeKing API

Quickstart

Base URL is https://vibeking.fun. Everything is a GET. There is nothing to configure:

curl -s "https://vibeking.fun/api/products?limit=5"

Press "Run it" to fetch a real response from this page.

Endpoints

GET /api/products — structured JSON

The one you want. Returns approved products as structured JSON, ordered by upvotes descending, wrapped in an envelope that also carries the directory total and the liveness aggregate.

{
  "success": true,
  "total": 2587,            // approved products in the directory
  "count": 1,                 // rows in this response
  "offset": 0,
  "liveness": { "live": 2408, "at_risk": 178, "dead": 0, "unchecked": 1 },
  "data": [
    {
      "id": 3527,
      "name": "APIDot",
      "tagline": "Unified AI API Platform for Image, Video, Music, and Chat Models",
      "description": "APIDot is a unified AI API platform built for developers ...",
      "url": "https://apidot.ai/",
      "maker": "APIDot",
      "ai_tool": "",
      "category": "AI Coding",
      "upvotes": 951,
      "created_at": "2026-01-12 02:20:34",
      "logo_data": "https://submitsaas.com/images/product-icon-ec57....jpg",
      "live_status": "live",
      "last_checked": "2026-08-05 12:00:54",
      "fail_count": 0
    }
  ]
}

Envelope fields: success, total (whole directory, not this page), count (rows returned), offset, liveness, and data. Each row also carries maker, maker_x, use_cases, audience, alternative_to and approved — the optional metadata fields are often empty strings or null, so code defensively.

Categories are data, not a fixed enum. Derive the current set from the dump rather than hardcoding it.

GET /api/list — paginated rendered rows

One page of the directory as a pre-rendered HTML fragment plus pagination metadata. This is what the site's own “load more” uses. Handy if you want VibeKing's row markup verbatim; use /api/products if you want fields.

curl -s "https://vibeking.fun/api/list?category=Game&page=1"

{
  "html": "\n    <div class=\"product-row\">...</div>\n",
  "count": 15,
  "page": 1,
  "total": 15,
  "hasMore": false
}

GET /api/search — keyword search

Matches name, tagline and description. Same response shape as /api/list, plus the echoed query. Also pre-rendered rows.

curl -s "https://vibeking.fun/api/search?q=todo"

{
  "html": "\n    <div class=\"product-row\">...</div>\n",
  "count": 5,
  "page": 1,
  "total": 5,
  "hasMore": false,
  "q": "todo"
}

Exact name matches rank first, then prefix matches, then everything else by upvotes. Nothing found is 200 with count: 0 — check count, not the status code.

GET /badge/{id}.svg — live badge

A live “Featured on VibeKing” SVG for any approved product, rendered from current data and cached for an hour. An unknown or unapproved id returns 404 — badges cannot be minted for products that are not really listed.

<a href="https://vibeking.fun/product/3527" target="_blank" rel="noopener">
  <img src="https://vibeking.fun/badge/3527.svg" alt="Featured on VibeKing" width="250" height="60">
</a>

The badge is the only endpoint without a CORS header — <img> embeds do not need one, but a cross-origin fetch() of the SVG will be blocked. Makers who display the badge on their own site unlock a followed link on their listing; see Verified Makers.

Liveness: the part nobody else publishes

Directories rot. Links break, projects get abandoned, domains lapse, and most directories keep serving the corpses because nobody checks. VibeKing re-probes every listed URL on a schedule and every /api/products response ships the aggregate:

"liveness": { "live": 2408, "at_risk": 178, "dead": 0, "unchecked": 1 }

Per-row you also get live_status (live, at_risk, dead, or null when it has not been probed yet), last_checked, and fail_count. Bot-blocking responses (401, 403, 405, 429, 451) are deliberately not counted as failures, because a site that blocks our probe is not a dead site.

If you are building anything on top of a directory — a dataset, a newsletter, a leaderboard — that field is the difference between real numbers and inherited rot.

Copy-paste examples

curl

# The whole envelope, first 5 products
curl -s "https://vibeking.fun/api/products?limit=5"

# Page 2 of the newest submissions (rendered rows)
curl -s "https://vibeking.fun/api/list?sort=new&page=2"

# Search
curl -s "https://vibeking.fun/api/search?q=todo"

# A product badge
curl -s "https://vibeking.fun/badge/3527.svg?theme=light&style=rank"

JavaScript (fetch)

// Top 10 vibe-coded products by upvotes. Runs in the browser too — CORS is open.
const res = await fetch('https://vibeking.fun/api/products?limit=10');
const { total, liveness, data } = await res.json();

console.log(`${total} products indexed, ${liveness.live} confirmed live`);
for (const p of data) {
  console.log(`${p.upvotes}  ${p.name} — ${p.url}`);
}

Python (requests)

import requests

# Page through the whole directory 500 at a time.
products, offset = [], 0
while True:
    r = requests.get("https://vibeking.fun/api/products",
                     params={"limit": 500, "offset": offset}, timeout=30)
    body = r.json()
    products += body["data"]
    offset += body["count"]
    if offset >= body["total"] or body["count"] == 0:
        break

print(len(products), "products")
print(sorted({p["category"] for p in products}))

Machine-readable specs live at /openapi.json, /.well-known/ai-plugin.json, /llms.txt and /llms-full.txt.

Open source

Two public repos back this API — docs and runnable clients in one, a regenerated dataset in the other.

Fair use

• Cache aggressively. ?limit=all is roughly 3 MB; once an hour is plenty.
• Page with limit + offset instead of pulling the whole dump on every request.
• Badge SVGs already carry Cache-Control: public, max-age=3600 — do not poll them.
• No published quota, but the site sits behind Cloudflare and abusive traffic gets blocked at the edge.
• Using the data is fine, including commercially. Credit the makers and link back to /product/{id}.

FAQ