Developers · Early access

The Cookie Hood API

Every score our AI engine produces — HoodScore, Mindshare, social authenticity, LLM-read sentiment, and evidence-carrying risk — as clean JSON. Built for traders, bots, and AI agents making decisions on Robinhood Chain. First keys roll out in the coming weeks; apply below.

1.5K

tokens tracked

144.9K

pools indexed

513

tokens under active risk flags

934

launches in the last 24h

01 · Endpoints

Live data, not a mockup

Every endpoint below is running today behind this site. Early access adds keys, stable versioning, and documented rate limits — not new promises. All scores are our own derived metrics, and their meaning is public on the methodology page.

GET /v1/leaderboard

Attention leaderboard

HoodScore and Mindshare rankings for the whole chain — real buyers, bot-filtered social attention, wash-discounted volume.

[{
  "symbol": "CASHCAT",
  "token": "0x9b7f…c4e2",
  "mindshare": 21.07,
  "delta_24h": 3.9,
  "organic_score": 74,
  "unique_buyers_h24": 8213,
  "sentiment": { "bullish_pct": 62.1,
    "bearish_pct": 11.4, "count": 419 },
  "risk_score": 10
}, …]

GET /v1/risk/{address}

Risk & sellability

Evidence-carrying flags — honeypot simulation verdicts, sell blockades, concentration, sybil clusters — with an open/resolve lifecycle.

{
  "risk_score": 50,
  "band": "high",
  "flags": [{
    "code": "HONEYPOT_SIM",
    "severity": 50,
    "evidence": { "buy": "ok",
      "sell": "reverted" },
    "detected_at": "2026-07-12T09:41:00Z"
  }]
}

GET /v1/social/{address}

Social intelligence

Mentions, impressions, LLM-read sentiment, bot and coordination share, top voices, and first-time KOL entries.

{
  "mentions": 312,
  "impressions": 1840000,
  "bot_share": 0.18,
  "coordinated_share": 0.07,
  "sentiment": { "bullish_pct": 58.3,
    "bearish_pct": 14.1, "net": 0.442 },
  "top_voices": [{ "handle": "…",
    "followers": 48200, "is_kol": true }]
}

GET /v1/divergence

Divergence alerts

02 · For AI agents

Agent-native by design

Honest nulls

Missing data is null, never a fake neutral. An agent acting on our scores never mistakes “not measured” for “average”.

Machine-checkable risk

Every risk flag ships the evidence that raised it, so an agent can apply its own threshold instead of trusting a black box.

Pay-per-call, no accounts

x402 pay-per-call and an MCP server are on the roadmap: agents pay per request in stablecoins and plug Cookie Hood straight into their stack.

03 · Early access

Apply for a key

Keys are issued in waves, in the order applications arrive. Live products and agents get priority. Access is free during early access with fair-use limits; paid tiers come later.