AI Agent Trust Glossary
Key terms and concepts in AI agent trust, the x402 protocol, and service reliability scoring.
AI Agent Trust Score
A numerical rating (0-100) that measures how reliable and trustworthy an AI agent service is before any transaction occurs.
Similar to a credit score, it aggregates multiple behavioral signals into a single actionable number. ScoutScore generates trust scores by monitoring services across four pillars: Contract Clarity, Availability, Response Fidelity, and Identity & Safety.
Agent Trust Infrastructure
The systems and protocols that enable AI agents to evaluate each other's trustworthiness before transacting.
ScoutScore is trust infrastructure for AI agents - it provides the scoring layer that sits between an agent deciding to pay and the actual payment. Without trust infrastructure, agents spend money blindly.
x402 Protocol
A protocol created by Coinbase that enables AI agents to make HTTP-based micropayments using USDC stablecoins.
It uses the HTTP 402 Payment Required status code as the payment trigger. When a service returns a 402 response, the paying agent sends USDC to complete the transaction. The x402 ecosystem has grown to 19,000+ cataloged endpoints.
Service Fidelity
A measure of whether an AI agent service actually delivers what it advertises.
ScoutScore's fidelity probes test services every 6 hours by sending real requests and comparing responses against advertised schemas and descriptions. The average fidelity score across the x402 ecosystem is 52 out of 100, meaning many services still fail to deliver on their promises.
Schema Phantom
A service that advertises an API schema (input/output definitions) in its metadata but fails to actually serve that schema when probed.
ScoutScore detects schema phantoms during fidelity checks. If a service claims to accept JSON input and return structured output but actually returns errors or empty responses, it gets flagged as SCHEMA_PHANTOM.
Wallet Spam Farm
A single cryptocurrency wallet address that registers hundreds or thousands of fake services on the x402 protocol.
The largest spam farm ScoutScore detected used one wallet to register 10,658 services, all with identical "Premium API Access" descriptions. ScoutScore applies a -25 to -50 score penalty for services associated with spam farm wallets.
Payment Facilitator
In the x402 ecosystem, the service that receives payment from an AI agent in exchange for performing work.
Payment facilitators range from legitimate AI services (image generation, data analysis) to spam farms that collect payments without delivering results.
Agent-to-Agent Payment
A transaction where one AI agent pays another AI agent or service to perform work, without human intervention.
The x402 protocol enables these payments using USDC. Trust scores help the paying agent decide whether to proceed.
ERC-8004
An Ethereum standard (currently in draft) for on-chain agent reputation.
It defines a protocol-level mechanism for recording and querying agent trust data. ScoutScore is registered as ERC-8004 agent #1308.
On-Chain Reputation
Trust and reputation data stored on a blockchain, making it tamper-resistant and publicly verifiable.
ScoutScore operates on Base and Solana, with trust scores anchored to on-chain identity.
Trust Layer
A middleware layer that sits between AI agents and the services they interact with, providing trust verification before transactions occur.
ScoutScore functions as a trust layer for the x402 ecosystem.
Contract Clarity
One of ScoutScore's four scoring pillars (weighted 20%). It measures how well a service defines its interface.
This pillar evaluates pricing accuracy, schema completeness, and metadata quality. Services with complete schemas and accurate pricing score higher.
Response Fidelity
One of ScoutScore's four scoring pillars (weighted 30%). It measures whether a service actually delivers what it promises.
Fidelity probes send real requests and verify the response matches advertised behavior. This is the most heavily weighted pillar because delivering on promises is the foundation of trust.
Payment Gate
A trust threshold that determines whether an AI agent should proceed with a payment.
For example, an agent might be configured to only pay services with a ScoutScore of 75 or higher (HIGH trust level). Services below the gate are blocked.
Behavioral Monitoring
The practice of continuously observing how AI agent services behave over time rather than relying on one-time audits.
ScoutScore runs health checks every 30 minutes and fidelity probes every 6 hours, building a behavioral profile that captures reliability trends.
Agent FICO Score
A colloquial term for AI agent trust scores, drawing an analogy to consumer FICO credit scores.
Just as FICO scores help lenders evaluate borrowers, agent trust scores help AI agents evaluate services before paying. ScoutScore positions itself as "the FICO score for AI agents."