Delivery Rate
% of paid calls that return a non-empty 2xx response. The bedrock signal: caught nine in ten ghost merchants in our backtest.
window 7d · weight 0.40
Repute scores every merchant on Arc in real-time, backed by Circle Developer-Controlled Wallets and the x402 payment protocol. Query a trust score in one line before your agent spends USDC. Catch ghosts, ruggers, and flaky nodes before they cost you.
x402 micropayments let any agent buy any service. But there's no trust layer: no Yelp, no PageRank, no credit bureau. Repute is the receipts-aware index that watches every payment on Arc and tells your agent who to trust before sending money.
A passive indexer on Arc. No SDK lock-in, no consensus rituals. Plug in, query, decide.
A live terminal for operators: streaming feed, merchant leaderboard, fraud alerts, and an API explorer. Free to access, no login required.
SDK for agent builders. CLI for operators and scripts. MCP for Claude Code — tools appear automatically, no glue code. All three sit on the same REST API underneath.
// 1. Install npm i @repute/sdk // 2. Query before every payment import { Repute } from '@repute/sdk'; const repute = new Repute({ baseUrl: 'https://repute.xyz' }); const score = await repute.score('0x15481D7B...'); if (score.verdict === 'SAFE_TO_PAY') { await circleTransferUsdc(walletId, score.address, score.price_per_call); } // Pick from the full leaderboard — safe merchants only const { merchants } = await repute.leaderboard(); const trusted = merchants.filter(m => m.trust_score > 75 && !m.fraud_flag); // Subscribe to fraud alerts via webhook await repute.subscribe('https://myagent.xyz/hooks', ['fraud_flag']);
# No install needed — just npx # Full trust profile for one merchant npx @repute/cli score 0x15481D7B... # Output: # PriceFeed Pro 0x15481D7B... # Verdict SAFE_TO_PAY # Trust ████████████████████ 88/100 # Reliability ████████████████████ 99.1% # Fraud flag none # Safe merchants only, top 5 npx @repute/cli board --safe --limit 5 # Score multiple addresses at once npx @repute/cli batch 0xABC... 0xDEF... 0x123... # Network stats npx @repute/cli stats # Point at any Repute API instance npx @repute/cli score 0x... --api https://repute.xyz
// Add to .claude/settings.json in your project { "mcpServers": { "repute": { "command": "node", "args": ["--experimental-strip-types", "./packages/mcp/src/index.ts"], "env": { "REPUTE_API_URL": "http://localhost:3001" } } } } // Claude can now call Repute tools natively — no glue code // Available tools: // repute_score — full verdict for one address // repute_batch — score up to 20 addresses // repute_leaderboard — sorted list, safe_only filter // repute_stats — network totals // repute_feed — recent indexed payments // repute_subscribe — register fraud alert webhook // Example — Claude routes payments automatically: // "Pay the best merchant for price data right now" // Claude: calls repute_leaderboard(safe_only: true) // picks top result, calls circleTransferUsdc()
# Score a merchant (x402-gated · $0.0001 USDC in live mode) curl https://repute.xyz/score/0x15481D7B... \ -H "X-Payment-Proof: <tx-hash>" # Response { "name": "PriceFeed Pro", "trust_score": 88, "verdict": "SAFE_TO_PAY", "reliability_pct": 99.1, "fraud_flag": null, "price_per_call": 0.0003 } # Free endpoints (no payment required) curl https://repute.xyz/leaderboard curl https://repute.xyz/stats curl https://repute.xyz/feed?limit=20 # Subscribe to fraud alerts curl -X POST https://repute.xyz/subscribe \ -H "Content-Type: application/json" \ -d '{"url":"https://myagent.xyz/hooks","events":["fraud_flag"]}'
Each merchant gets a composite score of 0–100 derived from five independent signals. All five are queryable. None are guesses.
Same x402 protocol your agent uses for everything else. No subscriptions, no seats. Just per-call settlement on Arc.