SmartPerps: paid x402 Hyperliquid smart-money positioning for trading agents (what the 120 most profitable Hyperliquid traders of the last 30 days hold right now, per coin: long/short trader counts and notional, net bias, 1h/4h/24h flows, entries, leverage, liquidation clusters, top positions, a market-wide scan and a live profile of any Hyperliquid account; snapshot every 5 minutes in a Durable Object, fail-closed, read-only analytics, not financial advice). Cloudflare Worker, Hono, x402 v2 + v1, Base / Polygon / Arbitrum One / Solana, remote and stdio MCP.
Is the smart money long or short? SmartPerps tracks a rules-based cohort of the 120 Hyperliquid accounts with the highest 30-day PnL (public leaderboard, all-time PnL > 0, account value >= $100k, market-maker-like volume excluded) and reads their open perp positions every 5 minutes. Read-only analytics from public data.
Not financial advice. Past performance does not guarantee future results. Bias labels are descriptive statistics of current public positions, not predictions or trade signals. SmartPerps never trades, never holds funds and never asks for keys.
https://api.youraitools.co/smartperps/mcp (tool smart_money_bias, free preview)node src/mcp-stdio.js (tools smart_money_positioning, smart_money_scan, hyperliquid_wallet_profile); npm package pendingdocker build -t smartperps-mcp . && docker run -i --rm smartperps-mcp| Route | Price | Returns |
|---|---|---|
GET/POST /v1/lite?symbol=BTC |
$0.003 | cohort traders long/short, long/short/net notional, bias, 1h/24h net flow, snapshot age, coverage |
GET/POST /v1/coin?symbol=ETH |
$0.02 | lite + 1h/4h/24h flows, avg entry vs mark, leverage, uPnL, concentration, liquidation price bands (count + notional per band), top 10 positions (addresses truncated, no liquidation prices) |
GET/POST /v1/scan?limit=10 |
$0.03 | cohort book + coins most net long / net short / one-sided, biggest 24h buying and selling |
GET/POST /v1/wallet?address=0x... |
$0.01 | live profile of the Hyperliquid account you supply: equity, leverage, positions with distance to liquidation (whole %), 30-day fills, cohort rank |
GET /v1/preview?symbol=BTC |
free | bias + trader counts (10/min per IP) |
Pay with x402 v2 (USDC on Base, Polygon, Arbitrum One or Solana) or v1 (Base), or a prepaid
YourAITools credits key (Authorization: Bearer yatc_...). Only HTTP 200 is
settled: bad input (400), unknown coins (404) and stale or incomplete data (503) are never charged.
/v1/lite, /v1/coin, /v1/scan, /v1/preview, MCP) never return full account addresses: top
positions show a truncated address (0x1234…abcd) and no liquidation price.25%) with a price range, position count and notional per band. Notional is withheld for bands with fewer than 3 positions, so no band maps to one account.
/v1/wallet profiles only the address the caller supplies, and shows each position's distance to liquidation in
whole percent instead of the liquidation price.unknown, confidence low and sources_failed.hyperliquid_leaderboard is listed.api.hyperliquid.xyz with fallback api-ui.hyperliquid.xyz.Cron (every 5 min) → stream-parse the public leaderboard (hourly) → cohort → clearinghouseState per account →
per-coin aggregates + 26 h history in a SQLite Durable Object. Paid requests read the snapshot: no upstream call
between the 402 and the answer. Usage counters live in a separate SQLite DO (no KV).
npm test (node --test), npm run gen:examples (re-record example outputs), wrangler deploy.
MIT. Built by agentgate-builder. Source: https://agentgate-source.pages.dev/x402-smartperps/
agentgate-source: code mirror on Cloudflare Pages. All repos