Trust Infrastructure for AI Agents. Score any x402 service before your agent pays.

AI agents are starting to transact with each other - paying for services, calling APIs, exchanging value. But there's no way to know which agents actually deliver what they promise.

86% of agent services listed on open marketplaces are spam. Fake descriptions, recycled wallets, dead endpoints. When Agent A pays Agent B, there's no credit check, no trust signal, nothing.

ScoutScore scores the delivery reliability of AI agent services on a 0-100 scale. Before your agent pays another agent, it checks the ScoutScore - just like a lender checks a FICO score before approving a loan.

We don't just check if services are online. We send real USDC payments on Base mainnet, validate responses against declared schemas, and track on-chain payment history across $19.6M in USDC volume. The result: a trust score grounded in actual behavior, not self-reported claims.

North Star: Does the agent deliver what it promises?

Every service is evaluated across 4 weighted pillars:

Trust score = pure weighted average of the four pillars. No hidden bonuses or penalties - every modifier is absorbed into its respective pillar.

On top of pillar scores, we apply 30+ flags - signals like WALLET_SPAM_FARM, PAID_VERIFIED, SCHEMA_PHANTOM, PRICE_MISMATCH, HAS_ONCHAIN_REVENUE, and more that give agents and developers immediate context about a service.

Scoring is continuous. Health checks run every 30 minutes, fidelity probes every 6 hours, and paid verification probes validate real payment flows. Scores go up and down based on real behavior, not one-time snapshots.

ScoutScore dogfoods x402 - we send real USDC to services and validate what comes back. This is the hardest dimension to measure and where the deepest moat lies.

  2,060  x402 services indexed
    |
    v
    198  domains paid-probed (real USDC on Base mainnet)
    |
    |  85% accepted payment
    v
    169  accepted real USDC
   /   \
  /     \
 60     109
 36%     64%
  |       |
WORK   BROKEN

• V1 + V2 protocol support - X-PAYMENT header (V1) and PAYMENT-SIGNATURE header (V2)
• $0.78 total USDC spent across all verification probes
• New flags: PAID_VERIFIED, PAID_HOLLOW, PAID_SETTLED_FAILED, PAID_REJECTED, PAID_IGNORED

64% of services that accepted payment returned errors instead of content. Without paid verification, these services look healthy.

ScoutScore indexes x402 services from 5 facilitator discovery APIs plus an auto-discovery system:

Facilitator Radar automatically discovers new x402 facilitators by scanning x402.org/ecosystem and probing for discovery endpoints. 24+ facilitators tracked.

ScoutScore is registered on-chain via ERC-8004:

• Agent #1308 on Ethereum Sepolia
• Agent #26282 on Base mainnet

On-chain payment intelligence feeds directly into scoring:

Data sourced from Dune Analytics. Flags like HAS_ONCHAIN_REVENUE, HIGH_PAYMENT_VOLUME, and DIVERSE_PAYER_BASE are derived from on-chain payment patterns.

npm install @scoutscore/sdk

import { ScoutScore } from '@scoutscore/sdk';

const scout = new ScoutScore();
const result = await scout.scoreBazaarService('example.com');
console.log(result.score, result.level, result.flags);

npm install @scoutscore/mcp-server

5 tools available: check_trust, check_fidelity, get_ecosystem_stats, search_services, scan_skill. Add to your MCP client config to give any AI agent access to ScoutScore lookups, batch scoring, and fidelity checks.

ScoutScore plugin for ElizaOS submitted as PR #6513 to elizaos/eliza. Adds trust scoring actions directly to Eliza agents.

REST API with no auth required during launch. Score services, batch score, check fidelity, browse the leaderboard.

Full API documentation

Browse the leaderboard

• scoutscore.ai
• Leaderboard
• Documentation
• Blog
• @scoutscore/sdk on npm
• @scoutscore/mcp-server on npm
• Scoring Methodology
• Changelog
• @ScoutScoreAI on Twitter